Subscribe now

TL;DR

• Full-time virtual school penetration could reach 7% in an unconstrained environment, but the regulatory-adjusted market likely sits closer to 5%. Applied to roughly 50M U.S. K–12 students, a 7% penetration rate implies 3.5M full-time virtual students. After excluding states unlikely to permit broad virtual school expansion, the realistic addressable market falls to 2.45M students, or roughly 3.5x–4.0x today’s estimated 600K–700K student base.

• Regulatory resistance likely caps virtual school enrollment growth for 30% of the US K-12 population. States unlikely to permit an expansion of enrollments include California, New York, Illinois, New Jersey, Washington, Massachusetts, Oregon, and Connecticut.

• Regulatory hurdles, not a lack of family demand, continue to serve as the primary bottleneck for industry growth. Families have signaled a clear desire for these programs post-pandemic, but restrictive authorization frameworks, oversight rules, moratoriums, and enrollment caps continue to block them.

• Growth in virtual enrollments does not automatically benefit education management organizations like Stride or Pearson’s Connections Academy. Each state structures virtual schooling differently. Florida, for example, built a district-centered virtual infrastructure around its own statewide entity, Florida Virtual School.

• The virtual school market likely continues impressive near-term growth, particularly with the loosening of rules in Texas. Recent legislative changes removed structural barriers that historically constrained virtual enrollment growth in one of the country’s largest K–12 markets.

• AI represents the great unknown, but likely creates further options for families and potentially redefines virtual instruction.

Context: Stride/K12 as the Virtual School Market Bellwether

This concludes my series on the virtual school market and Stride/K12 (LRN). Thus far I’ve written about:

• Stride as an investment, published in Seeking Alpha.

• AI’s impact on Stride / K12

• What happens to virtual schools that leave Stride / K12

As the market leader in virtual turnkey management services, Stride remains the central figure in any discussion regarding the industry’s position.

Stride put the industry in Wall Street’s spotlight back in October. The stock price declined by 50% following a meaningful challenge with Stride’s implementation of a new Canvas learning management system. This operational crisis served as a stark reminder of the Herculean effort to execute virtual school operations successfully. Savvy investors viewed this temporary dip as a buying opportunity for a quality business, especially since the stock had appreciated sixfold since the pandemic.

I’ve spoken with a number of investors in an informal (not professional) capacity over the past six months. For new readers to this newsletter, I served at competitor Connections Education / Pearson as head of strategy for six years including a stint for two of them as general manager of the $60M revenue generating institutional business selling virtual school solutions to school districts. Previously I served as a sell-side analyst covering the stock at several investment banks.

One of the first questions that investors asked when kicking the tires on the name was “how big can this market get”?

My response was usually along the lines of “I have no idea.”

This wasn’t an exercise in humility.

The Addressable Market Falls from 7% to 5% Penetration After Political and Regulatory Constraints

Prior to the pandemic, I used to say that the national addressable market was 3% of the K-12 student population. That full-time K-12 virtual instruction would always be a niche offering within the mosaic of broader K-12 solutions. At that time, in the states with the most progressive regulatory environment, penetration pre-pandemic appeared to peak at perhaps 3%.

For most pandemic related behaviors, consumers reverted back to the status quo.

Not for this market.

The pandemic increased awareness both of the category’s existence and also the capacity for families to benefit from the virtual education experience. Notice how I say family rather than students. Education is a family affair. Understanding this is central to understanding the opportunity with full-time virtual instruction.

I’ve had some time to reflect and research how virtual schools have fared post-pandemic, enough to attempt to answer the question of how big this market can get.

I believe the full-time virtual school penetration rate could eventually approach 7% of the U.S. K–12 population in an unconstrained environment. Applied to roughly 50M public K–12 students, that implies a theoretical market of 3.5M full-time virtual students.

But states dictate K-12 education policy. Some will not allow or support this modality. If roughly 30% of U.S. students live in states unlikely to permit large-scale cyber charter expansion, the regulatory-adjusted penetration rate falls to approximately 5%, or roughly 2.45M students. That compares with today’s estimated 600K–700K full-time virtual students.

Charter school law matters because it created the vehicle that allowed full-time virtual schools to scale. In states that authorized statewide cyber charters, families could leave their local district and enroll in a public online school that served students across the entire state. That gave virtual school operators scale, which allows them to invest further in their schools and market the offering to more and more students. States that allow cyber charters create a path for rapid virtual school growth. Districts can serve virtual enrollments but few have scaled meaningfully — unless they operate through a framework that grants them a statewide purview.

Unfortunately, this market lacks clean reportable data allowing for analysis. The best national estimate comes from NEPC’s 2023 Virtual Schools in the US report, which found that full-time virtual school enrollment increased from 332K students in 2019–20 to 644K students in 2020–21. Because national enrollment declined somewhat afterward but remained far above pre-pandemic levels, I use roughly 600–700K students as the current national market-size range.

Here’s how I come up with the long-term 5% penetration rate.

1) 50M students attend K-12 public schools.

2) Cyber charter schools in Pennsylvania now enroll 4% of the state’s K–12 population. Pennsylvania lacks some of the structural barriers present in other states and allows statewide cyber charter operators to enroll students. Including district programs, whose data it not readily available publicly, the total virtual penetration rate might be closer to 5%.

3) Pennsylvania cyber charter enrollment growth continues to accelerate post-pandemic. Assuming a decelerating growth rate over the next five years, a 7% penetration rate in that state seems plausible. Applying this rate to the national 50M student count, that would imply 3.5M full-time virtual students.

4) Policy and regulatory resistance likely reduces the effective national penetration rate from 7% to 5%. Roughly 30% of U.S. K–12 students, or about 15M students, live in states with significant political, regulatory, union, or institutional opposition to large-scale statewide cyber charter expansion. If those students remain largely inaccessible, then 30% of the theoretical 3.5M-student opportunity, or 1.05M students, should be excluded from the addressable market. That leaves a regulatory-adjusted opportunity of approximately 2.45M students.

5) The current national market has enough room to grow roughly 3.5x–4.0x. A 2.45M-student addressable market compares with today’s estimated 600K–700K full-time virtual student base.

6) Assuming a 20-year period to penetrate the addressable market, the industry could see perhaps 7% annual enrollment growth. Cell phones achieved mass-market maturity in 25 years. Landline phones took 80 years. Full-time virtual charter schools launched in the late 90’s. This category will almost certainly mature more slowly than cell phones. Whether it matures faster than landline phones remains unclear.

The Addressable Market Assuming no Regulatory or Policy Constraints: 7%

Pennsylvania provides the cleanest benchmark for estimating the long-term addressable market for full-time virtual schooling in the United States.

As you can see from the chart below, today’s cyber charter penetration rate of 4% represents double the 2% rate found pre-pandemic in 2020.

Can the penetration rate once again double over the next twenty years? We can decompose that question in terms of supply/demand:

• Supply - will the Pennsylvania governor/legislature seek to limit the offering as a greater % of students leave districts?

• Demand - are there enough families that need this option?

No expert exists today who can categorically answer those questions. As Yogi Berra astutely observed “it’s tough to make predictions, especially about the future.”

What gives me confidence on the likelihood of achieving this on the demand side is that Pennsylvania cyber charter enrollment growth continues to accelerate post-pandemic.

Acceleration implies that the flywheel is spinning faster. By flywheel I mean the self reinforcing loop where enrollments generate capital and awareness to drive further growth.

Parents continue shifting their kids into virtual options several years after schools reopened. From 2023 to 2026, enrollment grew from 57K to 69K students, a gain of nearly 12,000 students over three years.

My analysis assumes an analog of the Pennsylvania market to the national one. The reality is a bit more complicated.

Much of Pennsylvania’s cyber charter market concentration sits in one school. Commonwealth Charter Academy (CCA) enrolled close to 50% of the state’s cyber charter students. CCA has also expanded the definition of what a virtual school can offer families.

CCA has invested in physical and mobile infrastructure that has reduced core objections to virtual schooling: the lack of hands-on learning, socialization, and place-based student experience. The school operates Family Service Centers, mobile classrooms, AgWorks, and other experiential learning assets that bring parts of the physical school experience into a virtual model.

That infrastructure likely helped CCA enroll students who might otherwise have avoided the virtual school option.

Leaders at other virtual schools may seek to emulate CCA’s vision and execution. Or perhaps CCA serves as an outlier — Pennsylvania had unique funding characteristics (high funding per pupil) that allowed CCA to innovate.

Again there are fifty states, and each state’s circumstance is different. I could have reviewed another state with a high penetration rate.

The reason that I didn’t is that Pennsylvania has the cleanest data that can clearly tell the story. For example, Arizona appears to have one of the country’s highest online learning participation rates. But Arizona’s figure includes both part-time and full-time participation

Addressable Market Affected by Political and Regulatory Opposition

Some states will likely never fully support large-scale cyber charter expansion regardless of parent demand.

After full penetration of the addressable market we might see significant variability of rates in each US state like we see today.

DLAC, the industry leader in virtual school conferences/meetings, provides helpful data on full-time virtual enrollment and penetration rates. In its latest report it has a chart (see below) that shows wide variation in virtual school enrollment by state as of 2022–23.

Approximately 30% of the US K–12 population resides in states with deep-rooted political, regulatory, union, and institutional opposition toward statewide virtual charter schools. These states include:

• California

• New York

• Illinois

• New Jersey

• Washington

• Massachusetts

• Oregon

• Connecticut

These states differ in their exact policies, but they generally constrain statewide cyber charter growth through restrictive authorization frameworks, district-centered virtual models, enrollment caps, moratoriums, funding constraints, or heightened oversight requirements.

California has one of the largest number of full-time virtual enrollments of any state, with close to 40K enrollments. However, this represents only 0.65% of the full-time K-12 population in the state.

More importantly, conditions do not appear favorable for material enrollment growth in California. The state once was an early leader in virtual school adoption. Now not so much.

California imposed a moratorium on new non-classroom-based charter schools from January 1, 2020 through January 1, 2026. That moratorium has now expired. In 2025, the legislature passed SB 414, which would have created a tighter oversight framework for non-classroom-based charter schools, but Governor Newsom vetoed the bill. Big picture, California’s policy trajectory doesn’t look especially encouraging for statewide virtual or cyber charter schools.

If we’re looking at over a twenty year period, it’s conceivable that things may change for the better in terms of parental choice in these states.

Explaining Supply Side Constraints, What Limits Rapid Adoption of the Virtual School Option

Debates over virtual schools often start with a disagreement over what success means. Operators and families may define success as solving a specific student problem, while critics often focus on standardized academic outcomes, mobility, and district funding effects.

As the former CEO of Pearson’s Connections Education used to say, ultimately virtual schools solve problems. The intent isn’t to replace the traditional classroom. Not to offer superior academic outcomes. It’s not even supposed to serve as a long-term solution for every family. It’s a highly specific offering for a family to solve a specific need at a specific time.

If a student uses virtual school for less than a year to navigate a crisis and then return to their local district, has the offering succeeded?

Critics of the offering are rightfully concerned that transitioning between traditional brick-and-mortar and virtual can cause students to fall behind academically.

I bring my own bias to this issue. As the parent of a neurodivergent child who struggles with the noise and sensory environment of a brick-and-mortar school, I understand the category differently than I did when I analyzed it professionally. Families seek virtual schooling because failures in the classroom are truly a family crisis. It’s a win to just get the child to the classroom on-time in the morning.

Here are the traditional concerns that sector critics provide.

• Efficacy / outcomes / academic performance

• Funding diversion from districts

• Teacher staffing models

• Public / private partnerships, where for-profit education management organizations manage schools

In March the North Carolina Charter Review Board approved four new remote academies. Democratic North Carolina Sen. Jay Chaudhuri used the words “education malpractice” to describe the expansion of virtual charter schools.

Democratic Rep. Rodney Pierce said that “I just think this is part of a bigger overall agenda to weaken public education.”

In 2023 the National Education Policy Center (NEPC) published their annual report on the sector but issued a press release with the title “Virtual Schools Continue to Underperform, Have Little Research Support, and Lack Adequate Regulation”

The Network for Public Education stated that “No. Online charter schools, also called cyber schools and virtual schools, are a poor choice for students almost every time. They’re mostly a way for for-profit education operators to cash in by exploiting the most vulnerable families in the public education system.”

During the Biden Administration, the Government Accountability Office issued a report titled “K-12 Education: Department of Education Should Help States Address Student Testing Issues and Financial Risks Associated with Virtual Schools, Particularly Virtual Charter Schools”

Tennessee’s 2023-24 Virtual Education Report suggested that there are credible academic reasons to limit expansion, alluding to higher student mobility, weaker achievement, lower growth scores

These reports may bring up good points. Meanwhile, in Maryland we lack access to virtual and hybrid school solutions that might serve as the best option for my child.

Interestingly, policy advocates haven’t issued a major report over the past three years, at least one that I’ve seen. For all I know a piece might come out tomorrow.

I wanted to engage with critics of the offering ahead of this piece so I could understand the objections that they have post-pandemic. I reached out to various authors of these reports, sending them a link to my prior Substack article. I received no replies back.

My sense is that we’re not going to see a broad national deregulatory wave anytime soon. It’s conceivable that we’ll see a selective loosening of legislative/regulatory pressures as a step function. Like we’re seeing with Texas today. That’s how this sector grew over the past twenty years. That also explains why virtual enrollment growth will likely continue at a steady, sub-double-digit pace rather than through a broad national surge (at least absent another pandemic).

Laggard states will presumably at some point accept that virtual schooling belongs in the public-school toolkit. Likely with rules to control quality, mandates on hiring union teachers, and limits on how a for-profit management company can contract with a school.

Florida: A Case Study in How a Regulatory Structure May Not Materially Benefit Stride or Pearson

Florida demonstrates that the growth in full-time virtual enrollments in a given state may not necessarily accrue to education management organizations like Stride, Pearson, or other full-time virtual school operators.

Each state has its own structure for how it permits, regulates, and distributes online instruction. Pennsylvania created a statewide cyber charter ecosystem dominated by independent operators. Florida took a different path.

Florida launched Florida Virtual School (FLVS) in 1997 to expand course access for rural and underserved students. In 2000, the legislature transformed FLVS into an independent statewide school district. In 2003, Florida codified a performance-based funding model in which FLVS receives funding only after students successfully complete courses.

Florida did not build a Pennsylvania-style cyber charter market where large statewide operators compete directly against local districts for enrollment. Instead, Florida built a district-centered virtual learning system.

Florida law requires districts to provide virtual instruction options. Districts can:

• operate their own virtual programs,

• partner with FLVS (using FLVS teachers and operational capabilities),

• establish FLVS franchise programs (mostly FLVS courseware and tech systems),

• or contract with approved providers (external entity provides courseware and manages operations).

As a result, Florida created one of the country’s largest virtual learning ecosystems without producing a Pennsylvania-style cyber charter market.

In Florida, districts remain central to the administration of the virtual school experience.

Families can:

• remain enrolled in district schools while taking FLVS Flex courses (individual online course from FLVS without enrolling full-time in FLVS),

• participate in district virtual instruction programs,

• combine in-person and online schedules,

• enroll full-time in FLVS,

• or homeschool while relying heavily on FLVS coursework.

I estimate Florida’s effective virtual participation may already exceed 3%, although the state does not publish a clean figure that allows this to be measured precisely. Florida’s officially reported full-time virtual enrollment equals about 1.2% of statewide K–12 enrollment, but that figure excludes online instruction occurring through FLVS Flex, district channels, hybrid schedules, and homeschool pathways.

My estimate relies on FLVS disclosures showing that homeschool students represent 49% of the FLVS population (from the December 2025 “FLVS Home Education Guide”). Some of these students may function as de facto full-time virtual learners even though the state reports them under classifications tied to local districts rather than under full-time virtual enrollment categories.

Florida demonstrates that states can expand virtual learning substantially while still limiting the economic opportunity available to large independent virtual school operators.

Both Stride and Pearson operate within Florida. It’s just that the state’s own FLVS has substantial share that might otherwise have gone to Stride or Pearson.

Industry folks reading this piece might say that Florida is a unique circumstance driven by twenty years of policy, likely not to be replicated by other states. True, but all states have unique circumstances.

AI allows states to adopt the Florida model, as AI eliminates significant upfront costs for curriculum and technology development. There are a plethora of ex-Stride and Connections employees that can build out a FLVS for a state.

Laggard states with low virtual school penetration rates may follow the Pennsylvania model. They may follow the Florida model. They likely come up with another model. Which is to say, one cannot simply say that Stride and Connections have a 40% market share, and that as states allow this form of instruction Stride and Connections will maintain their market share. Maybe their market shares rises or maybe it falls with the addition of each new state.

Texas: The Largest Near-Term Growth Opportunity in Virtual Schools

Thus far, this piece has focused on the long-term size of the market. I’m adding this section because we may be witnessing a spike in virtual school enrollment in Texas given specific legislative changes. This will help test whether the addressable market remains more supply-constrained than demand-constrained.

Texas has 5.5M public K–12 students but only 60K full-time virtual enrollments, equal to a 1.1% penetration. So room exists between Texas and the Pennsylvania 4% cyber charter penetration rate. A Pennsylvania penetration rate of 4% would equal 220K full-time virtual students, more than triple the current enrollment level.

Historically, Texas constrained virtual enrollment through structural eligibility restrictions. Under the prior TXVSN framework, students needed to attend a Texas public school during the prior year in order to access full-time virtual programs. That rule blocked:

• incoming kindergarten students,

• students moving from out of state,

• students transitioning from homeschool or private school environments.

Governor Abbott signed Senate Bill 569 into law in May 2025, restructuring Texas virtual education policy. The legislation loosens prior structural restrictions and gives districts and charter schools more flexibility to operate and expand virtual programs. It doesn’t create unlimited immediate expansion. New full-time virtual campuses still require authorization.

Senator Paul Bettencourt, the bill’s author, stated he expects virtual enrollment in Texas to double by the 2028–29 school year. This would imply roughly 120K virtual students by 2029.

Texas will serve as a test for real market demand for states that restricted this type of instruction.

Texas will also represent a major test of whether future virtual expansion benefits national education management organizations.

Conclusion

No one can predict the ultimate penetration rate of full-time virtual schooling with precision. Too many variables remain unknowable: policy choices, operator execution, academic outcomes, AI adoption, district responses, and parent preferences.

Still, the current national market appears small relative to states that allow this form of instruction.

This market isn’t akin to the adoption curve of technologies like mobile phones.

It’s also worth noting that the line between virtual and brick-and-mortar instruction may blur over time as districts integrate the following into traditional schools:

• hybrid schedules

• AI-supported instruction

• asynchronous coursework

• remote tutoring

• flexible learning pathways

Political attitudes may also shift materially over the coming years. New research, operational successes/failures, student outcomes or cultural developments may either strengthen or weaken public support for virtual instruction. This may also affect public support for the education management organizations that operate virtual schools.

Like with other industries, AI may ultimately transform the category in ways that are difficult to comprehend today. AI-driven personalization, adaptive learning, tutoring and student support may improve the virtual learning experience. AI may also reduce barriers for districts and states seeking to build their own virtual infrastructure internally.

So here’s the range of outcomes as I see them.

Bull Case: Pennsylvania’s trajectory becomes the national model, and full-time virtual penetration ultimately approaches 7% in accessible states.

Base Case: Virtual schooling continues expanding, but policy resistance limits adoption in roughly 30% of the country. This produces a regulatory-adjusted national penetration rate closer to 5%.

Bear Case - High-profile efficacy studies, operational failures, fraud scandals, or political backlash trigger a regulatory pullback against virtual schools. States tighten enrollment caps, funding, authorizations, and accountability structures.

Whatever one thinks about policymakers, regulators, unions, or cyber charter operators with public company incentives, most participants in this debate genuinely believe they are acting in the best interests of students.

Virtual schools will not serve every student, and they don’t need to.

Gallup’s 2025 polling found that 23% of K–12 parents were either somewhat or completely dissatisfied with the education their oldest child receives. The pool of families experiencing dissatisfaction with their child’s current education is much larger than today’s full-time virtual enrollment base.

That’s a different way of sizing the potential market. If only a fraction of those dissatisfied families eventually choose full-time virtual schooling, the market can still grow several times larger than it is today.

Subscribe now

Disclosure

At the time of this writing, I hold a position in Stride/K12 (LRN) stock. I reserve the right to buy or sell shares at any time, in any amount, and for any reason without prior notice or subsequent updates to the readers of this newsletter. This analysis is for informational purposes only and does not constitute professional investment advice.

FAQ

How large could the U.S. K-12 virtual school market ultimately become?

The full-time virtual school penetration could approach 7% of the U.S. K–12 population absent regulatory and political constraints. That would imply 3.5M full-time virtual students nationally. After excluding states with onerous rules, the more realistic long-term addressable market likely sits at a 5% penetration rate with 2.45M students, representing roughly 3.5x–4.0x the current market size.

Why can K-12 virtual school enrollments triple from current levels?

Cyber charter penetration in Pennsylvania already equals 4% of statewide K–12 enrollment, and enrollment growth continues accelerating even after schools reopened post-pandemic. That suggests virtual schooling has become an acceptable offering for families with specific needs.

Why does Pennsylvania serve as an analog for the broader US market?

Pennsylvania lacks many of the structural restrictions present in other states and allows statewide cyber charter operators to scale broadly. Because of that, Pennsylvania provides one of the best real-world benchmarks for estimating what mature virtual school penetration may look like nationally.

What appears to be the primary constraint on virtual school adoption today?

Political and regulatory resistance continues limiting expansion through enrollment caps, restrictive authorization frameworks, moratoriums, district-centered systems, and heightened oversight rules.

Why do some states continue resisting virtual school expansion?

Policymakers, unions, and education advocates remain concerned about academic outcomes, funding diversion from districts, teacher staffing models, student mobility, and the role of for-profit education management organizations. Several states continue imposing enrollment caps, restrictive authorization frameworks, moratoriums, district-centered models, and heightened oversight rules.

Which states are most resistant to statewide cyber charter expansion?

California, New York, Illinois, New Jersey, Washington, Massachusetts, Oregon, and Connecticut represent states with significant political, regulatory, union, or institutional opposition toward large-scale statewide cyber charter growth. Together, these states represent roughly 30% of the U.S. K–12 population.

Why is Florida important to understanding the future of virtual schooling?

Florida demonstrates that states can expand online learning substantially without necessarily benefiting operators like Stride or Pearson’s Connections Academy. The state built a district-centered virtual learning system around Florida Virtual School (FLVS) rather than a Pennsylvania-style statewide cyber charter market.

How does Florida’s virtual school system differ from Pennsylvania’s?

Florida built a district-centered virtual learning system around Florida Virtual School (FLVS) rather than a Pennsylvania-style statewide cyber charter market. Districts retain substantially more operational control in Florida through FLVS partnerships, FLVS franchise programs, district virtual instruction programs, and hybrid models.

Does Texas represent a large opportunity for increased virtual school enrollments over the coming years?

Texas represents the clearest near-term growth opportunity in the sector. The state has approximately 5.5 million public K–12 students but only about 60,000 full-time virtual enrollments, equivalent to roughly 1.1% penetration.

What changed recently in Texas virtual school policy?

Texas passed Senate Bill 569 in 2025, loosening structural restrictions that historically constrained virtual enrollment growth. Prior rules required students to attend a Texas public school during the prior year before accessing full-time virtual programs.

How large are Stride and Pearson within the current market?

Stride and Pearson may collectively serve 40%–50% of the existing full-time virtual school market, depending on definitions. Stride currently serves roughly 250,000 virtual school enrollments. Pearson’s Connections Academy reportedly served roughly 100,000 students as of two years ago.

Why might AI improve virtual schooling over time?

AI may improve personalized instruction, tutoring, adaptive learning, asynchronous coursework, and student support systems. AI may also reduce barriers for districts and states seeking to build their own virtual infrastructure internally.

Could the distinction between virtual schools and traditional schools blur over time?

Yes. Over time, districts may increasingly integrate hybrid schedules, AI-supported instruction, asynchronous coursework, remote tutoring, and flexible learning pathways into traditional schools. The future may look less binary than today’s distinction between virtual and brick-and-mortar education.

What is the biggest long-term risk to the virtual school market?

The largest risk remains political and regulatory backlash. Negative efficacy studies, operational failures, fraud scandals, or broader political opposition could trigger tighter enrollment caps, funding restrictions, authorization rules, and accountability structures.

TL;DR

• This article examines what happens after virtual schools terminate their management contracts with Stride/K12. Literature historically focuses on drivers for schools leaving but ignores what happened afterwards.

• Across observable cases, schools face enrollment declines, execution issues, and financial pressure after separating from Stride.

• Enrollment loss is the most consistent outcome, not just in the following year but over the long-term. In multiple cases, students appear to shift to other K12-powered schools in the same state, suggesting family affiliation with the platform and curriculum, not the institution.

• Replacing the operating stack isn’t impossible but has proven challenging. Schools must rebuild curriculum, technology, and operations. Agora’s attempt ten years ago led to a $20M impairment charge for unusable computers and abandoned curriculum/technology development.

• Marketing and demand generation by Stride are underappreciated based on departing schools’ track record. Stride provides enrollment infrastructure and brand recognition that schools struggle to replicate after leaving. Marketing traditionally has not been a needed competence in K-12 education.

• Commonwealth Charter Academy (CCA) represents a successful transition from an EMO (Pearson’s Connections Education), but even it likely could not replicate its success in the current funding environment. While CCA in Pennsylvania is often used as a model for independence, it succeeded because of a specific high-funding state environment that is now under heavy government scrutiny.

• Bottom Line: There may be good reasons to leave an EMO, but there are many examples of challenges and few success stories.

While leading the strategy function at Pearson’s Connections Education, I believed that Education Management Organizations (EMOs) that serviced virtual schools would eventually make themselves obsolete. I saw that we would help a school reach a critical mass of enrollment and balance sheet surplus at which point the school board would inevitably choose to in-source capabilities. For the past several years, I assumed that the sizeable enrollment growth driven by favorable pandemic tailwinds would finally cause the customer attrition I had warned about for so many years.

While some customers have left over the years, Stride’s model has proven far more durable than I expected.

A decade ago, Stride generated less than $1B of revenue. Wall Street analysts now expect Stride to generate over $2.5B in fiscal 2026.

That disconnect between my hypothesis and reality raised my question: why aren’t more schools leaving?

I addressed this to some degree in my last piece regarding Stride’s positioning given the rise of artificial intelligence. Given the shift from clicking on links to getting synthesized answers from large language models (LLMs), the K12 brand in my opinion has greater value given how AI surfaces schools affiliated with the company and its curriculum.

But my analysis missed an important dimension, something I realized after conversations with former ex-Pearson colleagues.

No one focuses on what happens after a school leaves.

Do schools stabilize or struggle? Do enrollments hold or decline? Do expected financial benefits materialize? Do execution issues occur? Does leadership regret their decision to leave?

There are only a limited number of observable cases where schools have exited an EMO and where industry participants have evaluated outcomes through public data. This piece focuses on those cases. Not to critique those decisions, but to understand the operational and financial consequences that followed.

To be clear, each school operates under its own constraints. Leadership likely had valid reasons to pursue exists from EMO relationships. In the past these have included cost, control, service issues, academic outcomes, and governance.

This is a post-mortem based on observable outcomes: financial statements, enrollment trends, and public disclosures, including board materials where available.

The Structural Drivers of Financial Decline

Publicly available evidence suggests that when a school leaves Stride, the financial profile typically weakens, and not just in the first year. Based on my experience in the industry, this results from three specific operational headwinds:

1. Competition with Other K12 Schools in the State: Unlike traditional schools, virtual schools do not have a monopoly. Parents choose among options. When a school leaves the K12 ecosystem, it often competes directly against another K12-powered school in the same state. Families who value familiarity and perceived reliability may choose to remain within the K12 system.

2. Marketing and Enrollment Leadership: Stride operates a centralized scaled marketing engine that is extremely difficult to replicate at the individual school level. While schools can manage daily teaching operations and logistics, generating consistent enrollment is a different skill set that few in the K-12 world know. K12 executes this at a level most schools cannot match.

3. Operational Execution: Leaving an EMO requires replacing a fully integrated operating model. This includes curriculum, technology platforms, student information systems, vendor coordination, and staffing across multiple functions. Schools have to excel at each one of these functions. The basis of Stride’s success is that it can actually do so, with demonstrable success across multiple states and schools. In practice, individual schools struggle to execute across all of these functions at once, particularly during the transition period.

Why This Analysis Prioritizes Financials Over Academic Outcomes

This analysis focuses on financial and enrollment performance rather than academic outcomes.

That isn’t because academic outcomes are unimportant. It’s because evaluating efficacy in virtual schools is a far more complex challenge than one would think at first blush. By contrast, income statement performance and enrollment trends are observable, comparable, and easier to interpret across schools.

The complexity regarding evaluating academic outcomes relates largely to the non-traditional student population that virtual schools serve. Many families enroll after experiencing a material disruption that warrants a departure from a traditional school: academic struggles; bullying; health issues; attendance problems; or family instability. Traditional schools draw students from a geographic base but are otherwise a randomly assigned cohort. Virtual school students are a self-selected population, often already behind or disengaged at the time of entry.

That reality makes standard metrics difficult to interpret. Proficiency rates and graduation rates are level-based measures. They capture where students are, not how much progress they have made. For a population that starts behind, a school can improve outcomes meaningfully and still report weak proficiency. Differences in student composition further distort comparisons across schools.

For example, Agora Cyber Charter School, which altered its relationship with Stride / K12, serves a materially higher proportion of special education students. Special Education students represent 37% of total students vs. the 21% of public school students who receive services with IEPs in Pennsylvania. 1.89% of the student population is homeless. For Agora it’s 4.8%. Is it fair to evaluate Agora on traditional metrics like graduation or state test scores?

Conceptually, growth relative to a student’s starting point provides the correct framework for evaluating efficacy. But doing so requires student-level longitudinal data—prior achievement, baseline performance, and duration of enrollment—which isn’t available in public disclosures. Without that data, it’s difficult to isolate whether changes in outcomes reflect operator performance, curriculum quality, or shifts in student mix.

The US Government Accountability Office (GAO) in its 2022 report supports my perspective. The GAO found that full-time virtual schools enroll higher proportions of at-risk students and that available outcome data doesn’t adequately control for those differences. As a result, comparisons to traditional public schools and conclusions about school performance become difficult to interpret. States continue to rely heavily on proficiency and graduation rates, but those measures do not capture student starting points or progress. In practice, even formal accountability frameworks do not allow anyone to isolate whether outcomes reflect the operator, the curriculum, or the underlying student population.

From a regulatory standpoint, what ultimately matters is whether a school meets state accountability thresholds and avoids closure. That is a different question from isolating the impact of a change in operating model.

For that reason, this analysis focuses on financial performance and enrollment trends. While imperfect, these metrics provide a clearer and more comparable signal of how transitions away from EMOs have actually played out.

Case Study: Insight PA Cyber School

Insight PA provides the most recent example of a school in-sourcing capabilities.

The school launched in the 2020–21 academic year as Stride’s managed school in Pennsylvania, several years after Pennsylvania’s Agora Cyber Charter School reduced its dependence on Stride. The board terminated its relationship with Stride altogether for the 2025–26 academic year.

The timing of the divorce worked against them.

The transition to independence coincided with a reset in Pennsylvania’s funding framework for virtual schools. In November 2025, the state reduced funding by an estimated $178M, followed by a proposed additional reduction of roughly $75M in early 2026. Pennsylvania had historically been one of the most favorable funding environments for virtual schools, so some level of correction was inevitable.

In the first year operating without Stride, Insight saw a decline of 575 October enrollments.

The latest board material from Insight shows management projecting an operating loss of $10M for the year driven by lower than expected enrollments and the revised state funding profile. Cash declined from roughly $25.9M as of June 30, 2025 to $14.7M by March 31, 2026, with an expected year-end balance of around $12M.

In its 2023–24 budget, Insight management outlined an aspiration to move certain functions in-house over a one- to three-year period. That included hiring a chief technology and information officer, system analysts, recruiters, enrollment liaisons, and registrars.

The board materials do not identify a single driver for the transition. They repeatedly reference the school’s large fund balance. Management planned to deploy that balance for strategic investments while reducing its size, noting that regulators and the media could scrutinize it. That scrutiny has since materialized across Pennsylvania.

What stands out to me as a corporate finance professional are the assumptions that management provided in their 2025-2026 budget preview. Management projected $16M of cost savings, $8M of which would come from terminating the relationship with K12. The materials don’t explain where the other $8M comes from, which I find curious. That projection didn’t assume a decline in revenue.

The revenue assumption looks aggressive.

The financial case for leaving Stride looks attractive if costs fall and revenue holds. That framework ignores Stride’s role in demand generation. The K12 brand and marketing infrastructure drive enrollment, and schools cannot easily replace them. Assuming stable enrollment in the first year after separation ignores what other case studies show.

Across the examples reviewed in this piece, schools that leave Stride tend to experience enrollment declines. That pattern suggests the more conservative starting point is the opposite: assume disruption, not stability.

The intended use of the fund balance also becomes relevant in hindsight. Rather than being deployed purely for strategic investment, it has effectively served as a buffer against enrollment declines and funding pressure. As of March 31, the fund balance had declined to $14M.

The broader takeaway is that school stakeholders underappreciate the value of Stride’s platform. Families appear to have an affinity not just for the school, but for K12. When a school removes that affiliation, students do not necessarily stay.

Case Study: Agora Cyber Charter School

Agora Cyber Charter School launched in 2005 and scaled rapidly. Within less than a decade, enrollment exceeded 10,000 students by 2014, making it one of the largest virtual schools in the country.

In 2014, Agora’s board voted to alter its relationship with Stride. That decision included contracting with new vendors for core capabilities like student information systems, learning management systems, and hardware by 2015. Agora attempted to unbundle and replace a fully integrated operating model.

At the time, Agora stood as Stride’s largest customer. In fiscal 2015, it represented 14% of Stride’s revenue, contributing $129.8M. When the relationship changed, that revenue declined to $18.5M the following year as Agora moved away from a fully managed model and retained only certain services, including curriculum. Stride’s revenue declined from $948.3M in 2015 to $872.7M in 2016, and operating income fell from roughly $18M to $14M. The stock price reflected that disruption, declining from around $40/share to closer to $10/share.

After separating from Stride, Agora lost nearly half its enrollment. The pandemic temporarily boosted student counts, but the long-term trend remained negative. Some might attribute this to a more aggressive competitive landscape. A look at Commonwealth Connections Academy (CCA), a former partner of Connections Education in the same state, offers a telling contrast. Despite also choosing independence from its EMO and starting with a similar base of 8,768 students in the 2014/2015 school year (roughly 500 fewer than Agora), CCA has since grown its enrollment to nearly 35,000.

From Agora’s perspective, the financial trajectory also deteriorated following the transition. Based on IRS Form 990 data, revenue declined from approximately $125M in 2015 to around $95M in the subsequent pre-pandemic years. Revenue recovered during the pandemic to roughly prior levels but has since declined again to approximately $112M.

The school’s attempted operational transition circa 2015 saw challenges.

Agora partnered with Spider Learning to develop curriculum and with Education Elements to build its platform. Those efforts failed. The school disclosed settlement agreements with both vendors in its 2016 financial statements.

By 2017, Agora’s financial statements indicate that the school had reverted back to using K12 for educational products and administrative and technology services.

The financial consequences of the attempted transition saw Agora record a $10M impairment related to computer equipment acquired under a capital lease that was ultimately deemed unusable. In addition, the school recognized roughly $10M of impairment associated with abandoned curriculum, software, and platform development.

Operating cash flow declined from approximately $30M in 2016 to roughly $2M in 2017.

I haven’t been able to find the latest agreement between Agora and Stride, as to what services that Agora uses. The school still appears on K12’s marketing materials as the only public school in Pennsylvania affiliated with K12.

This case highlights the scale of execution risk. Agora tried to replace an integrated system with internal development and third-party vendors. That effort proved more complex, more expensive, and more disruptive than expected.

Case Study: Nevada Virtual Charter School (formerly Nevada Virtual Academy)

Nevada Virtual Academy launched in 2007 and operated for years as a Stride managed school. In 2023 the school moved to amend its charter to non-renew its contract with Stride and instead entered a contract with ACCEL for platform and curriculum services. ACCEL was co-founded by Ron Packard, the former CEO and founder of K12.

The board made this decision due to “service and financial concerns with Stride, K12”.

In a contract document explaining the change,

The current EMO has been unable to fulfill service agreements, has caused the school to receive multiple Notices of Concern due to finance mistakes, and was unwilling to be a partner but instead wanted to manage the school. The Board worked closely with SPCSA staff to understand the current and appropriate requirements of an EMO, appropriate pricing structures, and role an EMO should play within the school. Stride, K12’s business/operating model did not support the requirements, and in addition to the failures, the Board decided to non-renew their contract.

As with other cases, the transition itself introduced disruption. After terminating the contract, the school changed its name because Stride retained the rights to the prior name and acronym. There were also claims that K12 contacted families and notified them that they had been withdrawn from the school. While I cannot independently verify the specifics, the school attributes part of the disruption to communication issues during the transition.

The school estimates that it lost approximately 500 students who had previously indicated they intended to return for the 2023–24 academic year.

Enrollment has declined since the transition. State October enrollment reports show a drop of 750 students through the current academic year. The school attributes part of that decline to a charter change that limited enrollment to Clark County, as well as to broader statewide enrollment pressure.

The financial data does not reflect this enrollment decline linearly. Revenue was $18M in 2023, increased to $23.7M in 2024, and then declined back to $18.8M in 2025. The school’s fund balance increased materially, rising from $3.5M in 2023 to $14.5M in 2024.

I can’t reconcile the financial and enrollment data, so I’m reluctant to draw conclusions given my uncertainty about the data. On its surface, revenue and fund balance improved in the near term, even as enrollment declined.

At minimum, the transition created enrollment pressure without producing sustained financial growth.

Regarding academic outcomes, the high school showed some improvement under the state accountability framework, while the middle school continued to receive a “does not meet standard” designation. Given the enrollment decline and multiple simultaneous changes, the data does not support a clear conclusion about the impact of leaving Stride.

Case Study: South Carolina Virtual Charter School (SCVCS)

South Carolina Virtual Charter School launched in 2008 and ended its partnership with Stride Inc. (K12) on June 24, 2022.

The transition created immediate challenges.

Around the time of the separation, SCVCS posted on Facebook to address confusion among families who had received emails from K12 indicating that students had been withdrawn from the K12 system. The school needed to clarify enrollment status during the transition, which suggests communication breakdowns at a critical point in the cycle.

Enrollment data shows clear pressure following the separation. SCVCS lost roughly 3,000 students from its pandemic peak. At the same time, K12 continued to operate Cyber Academy of South Carolina in the same market.

In 2021, combined enrollment across both schools totaled approximately 10,000 students. By October 2026, that figure had declined to roughly 6,000. The K12-powered Cyber Academy lost approximately 1,000 students over that period, while SCVCS lost closer to 3,000.

That divergence suggests share loss rather than a neutral transition. SCVCS exited the K12 ecosystem but continued to compete against the same curriculum, platform, and brand through a remaining K12-powered school. Families who valued that ecosystem had a direct alternative.

Based on IRS Form 990 data, SCVCS generated approximately $37M of revenue in 2022, declining to roughly $26M by 2024.

This case highlights a structural challenge. When a K12-powered school remains in the same state, the exiting school competes against the platform it just left. The enrollment and revenue trends at SCVCS suggest that this dynamic can drive meaningful share loss.

Case Study: Oregon Virtual Academy / Evergreen Virtual Academy

Oregon Virtual Academy launched in 2008 and rebranded as Evergreen Virtual Academy in 2023, reflecting a shift away from the K12 model. The school has not provided a clear public explanation for the transition. Because it operates within a public school district, disclosures remain limited, and most available data comes from IRS Form 990 filings rather than detailed board materials.

The financials follow the same pattern seen in other case studies. Evergreen generated $32.6M in revenue at the pandemic peak in 2021 and then declined to approximately $14.0M by 2025, effectively reversing the entire COVID-era surge.

Enrollment trends moved in the same direction. Evergreen lost students after the transition, while Cascade Virtual Academy—which remains K12-powered—added approximately 739 students over the same period.

Evergreen exited into a market where a K12-powered alternative continued to operate with the same curriculum, platform, and distribution infrastructure. The enrollment divergence points to share loss. Some families likely chose continuity within the K12 system rather than remain with the standalone school.

Case Study: Commonwealth Connections Academy (CCA), the Aspirational Model

Commonwealth Charter Academy, formerly known as Commonwealth Connections Academy, represents the aspirational case of what a virtual school can become after leaving an EMO. CCA previously partnered with Connections Education and has operated independently for nearly a decade.

Revenue scaled from roughly $50M in 2011 to over $500M by 2024, with the most significant acceleration occurring during and after the pandemic. That growth translated into substantial balance sheet strength. Change in net assets turned sharply positive beginning in 2020, peaking at over $120M in 2022, indicating meaningful surplus generation rather than just top-line expansion.

CCA didn’t replace one EMO with another. It built its own operating model. That includes Edio, a custom platform developed for CCA. It has a fully controlled curriculum with more than 600 courses. It also expanded beyond a pure virtual model into physical and experiential infrastructure. Programs include:

• AgWorks – facilities that support controlled-environment agriculture, including aquaponics, hydroponics, and aeroponics

• TechWorks - programs focused on robotics, cybersecurity, computer science, drones, and advanced manufacturing

CCA built these programs to integrate career pathways, hands-on learning, and workforce preparation, supported by dual enrollment and external partnerships.

Other aspects of the school:

• They offer over 600 educational and social field trips across Pennsylvania.

• They have fully equipped mobile classrooms. These are trailer sized vehicles retrofitted with smartboards and science labs to allow hands on learning and face to face interaction.

  

However, the Pennsylvania Auditor General’s February 2025 report complicates the favorable narrative of CCA. The report highlights that cyber charter schools, including CCA, accumulated large surpluses under the state’s funding formula and deployed capital into significant non-instructional investments. Specifically, CCA spent close to $200M acquiring and renovating 21 buildings, raising questions about capital allocation and public funding levels.

The report argues that the existing funding model allowed cyber charters to retain excess funds rather than return them to districts.

So here’s the takeaway.

CCA’s performance reflects a combination of leadership, funding, and structure.

Leadership: In practice, school performance tracks in lock step with the quality of the school head and board. CCA succeeded because the leadership team not only had a vision but had the capability to execute on that vision.

Funding: CCA operated within a high funding environment that gave management the ability to invest in systems, infrastructure, and program development.

Structure: CCA operates within a single state. That removes the need to maintain fidelity across multiple jurisdictions and eliminates the constraints of a national operating model. Management does not have to standardize across dozens of regulatory regimes, which creates room to innovate.

CCA uses that flexibility to build programs tailored to Pennsylvania. It could develop hybrid models that could incorporate physical infrastructure and experiential learning. That’s something more difficult to execute within a national platform constrained by consistency requirements. A multi-state operator like Stride or Connections prioritizes uniform implementation. This provides advantages to its customers who benefit from best practices and operational optimization. But, this also limits how far any single school can deviate.

That local focus allows the school to design offerings that better match demand and differentiate its model.

CCA demonstrates that leaving an EMO can work, but under a very specific regulatory and funding environment that is now changing. The Pennsylvania government began to scrutinize cyber charter funding and spending in part because of CCA’s performance.

Paradoxically, the very success of CCA — its scale, surplus generation, and capital deployment — may make it harder for future schools to replicate the model.

Conclusion

Schools have valid reasons to leave Stride. Based on conventional statewide academic metrics, virtual schools often lag state averages for reasons discussed in this piece. Boards may perceive their EMOs as rigid because they operate through a standardized model. Boards and school leadership may conclude that they can operate more efficiently or deliver a better student experience by taking control of curriculum and operations.

The case studies point in a different direction. Schools struggle more than expected after they leave. The assumption that an EMO can be easily replaced often proves too optimistic. Schools face execution challenges and lose enrollment during the transition.

Independent service providers target these schools and promote alternatives to the EMO model. That dynamic creates options for EMO customers. Which is fine, all the power to them.

But school leaders and boards should recognize the mixed track record of leaving EMOs.

The financial case for leaving frequently relies on modeled cost savings. Those models should also assume revenue pressure. The key question isn’t whether disruption occurs. Rather, it’s the severity in the first year.

Schools must rebuild curriculum, technology, marketing, and operations at the same time. Each function may look manageable on its own, but coordinating them creates complexity. Stride’s integrated model works because it operates at scale. Schools lose that advantage when they leave.

Schools also underestimate marketing and demand generation. The K12 brand carries more weight than many boards assume. When schools remove that affiliation, they must rebuild demand while competing against other K12-powered schools in the same market. In several cases, students shifted to those remaining options.

Deficit protection adds another layer of value. Stride adjusts pricing to keep partner schools financially viable. Schools lose that flexibility when they leave, and the impact becomes visible when enrollment declines or funding changes.

Taken together, schools that leave should expect lower enrollment, lower revenue, and greater volatility in the near term. They take on full responsibility for execution and lose the stabilizing elements embedded in the EMO relationship.

The termination of a contract creates a difficult separation, with both sides acting in their own interest. That process can be disruptive for students and families in ways that directly impact retention.

Leaving Stride isn’t just a strategic decision. It’s a full operational reset. And based on the evidence we have today, the evidence suggests most schools underestimate how difficult that reset will be.

Disclaimer

I currently hold a position in Stride Inc. (LRN) and may buy or sell shares at any time without notice.

This analysis is based entirely on publicly available information, including company filings, state reports, and third-party data sources. While I have made a good-faith effort to ensure accuracy, I cannot guarantee that all information is complete, current, or free of error. Some interpretations may be incorrect.

This piece reflects my personal views and should not be construed as investment advice or a recommendation to buy or sell any securities. Readers should conduct their own due diligence and consult appropriate advisors before making any decisions.

While I authored this piece, I relied on AI to provide edits.

FAQ: Virtual Schools, EMOs, and Leaving Stride/K12

What happens when a virtual school leaves Stride / K12?

Schools typically face disruption. Across observable cases, schools that leave Stride often lose enrollment, encounter execution challenges, and experience financial pressure in the first year after separation.

Why do schools choose to leave an EMO like Stride or Connections Education?

Boards usually cite control, cost, and flexibility. They believe they can operate more efficiently, customize curriculum, or improve the student experience by bringing capabilities in-house.

Do schools save money after leaving an EMO?

They can reduce direct service costs, but savings often come with trade-offs. Many schools face revenue pressure due to enrollment declines, which can offset or exceed cost savings.

Why do enrollments decline after leaving K12?

Families often associate value with the K12 platform—its curriculum, technology, and brand—not just the individual school. When a school exits the ecosystem, some students shift to other K12-powered schools in the same state.

Is it difficult for a school to replace Stride’s services?

Yes. Schools must rebuild multiple functions at once: curriculum, learning platforms, student information systems, marketing, and operations. Coordinating these pieces creates execution risk.

How important is marketing in virtual education?

Marketing and demand generation are critical. Stride operates a scaled, centralized enrollment engine that most individual schools struggle to replicate. This often becomes a key driver of enrollment declines post-transition.

What is deficit protection, and why does it matter?

Stride’s model includes reducing its fees up until the point that its customer posts a budget surplus. Schools lose that support when they leave, increasing financial risk.

Are there successful examples of schools leaving EMOs?

Yes, but they are limited. Commonwealth Charter Academy (CCA) represents a notable success, but it benefited from existing scale and a favorable funding environment that may not be replicable.

Do academic outcomes improve after leaving an EMO?

The data does not allow for a clear conclusion. Virtual schools serve non-traditional student populations, and available metrics do not isolate whether changes in outcomes result from the operator, curriculum, or student mix.

What is the biggest risk when leaving Stride/K12?

Enrollment loss. Schools often underestimate how much demand depends on the K12 ecosystem and overestimate their ability to maintain stable enrollment during the transition.

Should schools leave an EMO?

Schools may have valid reasons to leave, but they should understand that it is a massive undertaking. It is a full operational reset with real execution and financial risk.

• Overall, AI is a net positive for Stride. The benefits on the demand side —stronger brand visibility and improved discovery — likely outweigh potential long-term risks from curriculum commoditization.

• AI strengthens Stride’s K12 brand. LLM-driven discovery surfaces schools affiliated with the K12 brand and thus improves customer visibility on the web.

• AI may erode the value of Stride’s existing curriculum over time. The cost of content creation is collapsing. What was once a core differentiator can increasingly be replicated at low cost, creating a long-term substitution risk.

• Curriculum commoditization could drive pricing compression if not managed. If schools can build their own curriculum using AI tools, they gain negotiating leverage.

• Stride needs to invest in next-generation curriculum to stay ahead. Stride’s customers have a greater likelihood of renewing at the current fee level if Stride establishes a clear product roadmap explaining what customers will receive. The key question is whether Stride can build an adaptive AI-driven curriculum fast enough to maintain differentiation before any potential pricing pressure.

• The durability of Stride’s business model likely remains intact in the age of AI. Stride has a moat. AI can strengthen the moat.

In January, I published an article on Seeking Alpha explaining my thesis as to why investors might want to buy Stride (LRN) stock. At the time, Stride’s challenged implementation of the Canvas Learning Management System (LMS) caused the stock to decline by over 50%. Would Stride’s failure catalyze enrollment churn? Would irate virtual school customers seek to exit their relationship with Stride?

My analysis was grounded in primary research: I reached out to families, scoured parent social media groups, and sat through virtual school board meetings. This is the sort of tedious work that some investors can do, basing an investment thesis on real facts. It led me to the conclusion that Stride’s operational storm had passed and the business had stabilized. Since then, the stock has appreciated by 34%, handily outperforming the S&P 500’s 3% return over the same period.

While the Q2 earnings report put the LMS operational concerns to rest, the current investment conversation on the stock has shifted to a broader and more consequential question: the transformative effects of AI. One cannot conduct channel checks for this sort of thing. Investors are now burdened with speculating about how AI will shift societal norms. How will students learn? How will teachers teach?

Every conversation I have with other investors on LRN inevitably lands on the same question:

How will AI affect Stride — does it still have a durable business model?

Amidst the SaaSpocalypse, a period where AI disruption fears triggered a massive sell-off in software stocks, the question of durability has become the only question worth asking.

What’s being asked of investors requires imagination.

My conclusion: AI could prove a net positive for Stride. But it also carries significant risk if the company doesn’t accelerate investment in curriculum.

While not an existential crisis on par with what some software vendors are facing, AI has eroded the value of Stride’s existing proprietary curriculum. This is definitionally correct. In its last fiscal year Stride invested $22M in curriculum development. Over the past three years Stride invested close to $60M in curriculum development. Given AI tools, how much capital would be required to build the same assets? Probably less.

It’s a leap though to conclude that lower curriculum asset value translates directly into lower enterprise value. Stride is not like edtech provider Chegg that saw a multibillion dollar business collapse over a two year period. Chegg’s end customer was the student, and those users could immediately switch to free alternatives like Google AI Overviews for “homework help”.

Stride operates differently. It signs long-term contracts with virtual schools that buy a bundled solution. Curriculum is only one piece of that bundle. More importantly though, customers aren’t just buying curriculum, they’re buying a system of instruction delivery.

To maintain the company’s market leadership, management must invest further in curriculum to sustain its competitive edge. Failure to do so may result in pricing pressure and customer attrition.

This piece is not a buy or sell recommendation on LRN stock. It is an attempt to answer a specific question: how does AI affect this business? Draw your own conclusions.

I remain long the stock. If I thought AI served as a concern for the next quarter, I would have already sold it.

My Vantage Point: Twenty Years with Stride and the K-12 Virtual School Industry

For readers unacquainted with my background, I was one of the first sell-side analysts to initiate coverage on Stride (then K12 Inc.) when it went public close to twenty years ago.

That work led me to Stride’s chief competitor, Connections Education (owned by Pearson), where I led strategy and later served as general manager of Connections’ institutional business.

My understanding of this company and this industry isn’t guided by financial statements or the curated management commentary found in earnings transcripts. My perspective comes from being in the room when operational decisions were made that moved the needle.

I’ve seen firsthand why boards choose to sign with an Education Management Organization (EMO) like Connections or Stride/K12, and more importantly, why they choose to leave.

I also have a deep appreciation for the historical difficulty of creating curriculum at scale. Building a comprehensive K-12 catalog is a massive undertaking that requires flawless execution. It served as a barrier to entry that has stood for decades. It remains to be seen if AI now threatens to dismantle that barrier.

Why School Boards and Customers Stay With Stride

To understand AI’s impact on Stride, it would be helpful to define the company’s often misunderstood relationship with its customers.

Stride operates primarily as a virtual school provider, servicing both traditional school districts and independent virtual charter schools under its flagship consumer brand, K12.

The student is the end-user.

But the real customer, the entity that signs the multi-year management contract, is the school board or the local district. Stride provides the infrastructure, the teachers, and the curriculum. The board retains governance and accountability.

For years, I believed this model contained a structural flaw: that once a managed school reached scale, it would internalize operations and exit the Stride partnership. That thesis has not played out. Enrollment growth and long contract durations suggest switching is far less common than I expected.

Reasons why schools stay with Stride:

• Many schools are financially constrained from switching. Most states provide insufficient per-pupil funding for virtual schools to afford contracting with a provider outside of Stride. Schools often carry a fund deficit that requires Stride to discount its pricing until the school builds a fund balance. Most do not have the balance sheet to in-source.

• Execution risk to replace Stride is high. Stride provides a turnkey, compliant, and proven operating model. Replacing that system is complex and time-consuming. It requires building or licensing curriculum, hiring staff, implementing systems, and ensuring regulatory compliance. Most boards are not incentivized to take on that level of operational risk. They are stewards, not disruptors.

• Families have loyalty to the K12 brand rather than the school. Stride typically operates several K12-powered schools within a single state. It turns out that parent loyalty does not reside with the local board or administrative authority. It resides with the K12 curriculum. If a specific board attempted to exit Stride, families could and would migrate to another K12-affiliated school in the same state to preserve continuity.

Customer attrition does occur. But when it happens, it’s typically in environments where funding is unusually favorable.

Pennsylvania is a clear example. The state has historically provided among the highest per-pupil funding levels for virtual schools, making independence more viable. Several Pennsylvania schools have exited EMO relationships as a result.

Insight PA Cyber Charter School provides a recent case study. The school exited its relationship with Stride last year.

In its first year independent of Stride, Insight as of October 2025 was 600 students below budget. An enrollment shortfall in itself is a serious issue, but this one coincided in the same year as when Governor Shapiro decided to cut funding for virtual schools in PA. As a consequence, Insight will see a close to $10M deficit. Unfortunately, the challenges required Insight to lay off staff. Insight has not been unique in facing these challenges, other virtual schools in Pennsylvania have laid off staff as well. But had Insight stayed in its partnership, Stride would have reduced its fees to Insight up until the point that Insight maintained a fund balance.

The take-away is that leaving the Stride partnership introduces material financial risk perhaps not acknowledged by stakeholders.

It’s difficult for an external observer to isolate the exact cause of the Insight enrollment shortfall. But one plausible factor is the loss of association with the K12 brand and curriculum.

From a parent’s perspective, the curriculum is often the product. In many cases, K12 is the school.

That dynamic likely explains why families remained with K12-powered schools when the company transitioned to the Canvas LMS.

The type of families enrolling their children at a K12 powered school are trying to fix a problem unsolved by the traditional school system.

For them, K12 isn’t just a consumer product.

It’s manna from heaven.

AI Benefits: Discovery as a New Moat

AI offers three distinct strategic advantages to Stride.

• Margin Expansion. Automation of labor-intensive tasks—enrollment, document verification, customer support, back-office support—creates a clear path to leaner operations and improved operating leverage. I’m not sure how big of an opportunity that is though, because management has already invested in these areas over the past several years.

• Improved Instructional Efficacy. With the cost of software and curriculum development meaningfully reduced, Stride can create more tools that measurably improve student outcomes. For example AI-enabled tutoring (now in beta mode at Stride) that provides real-time explanations, step-by-step guidance, and reinforcement tailored to each student’s level of understanding, creating a feedback loop between instruction and comprehension.

• Strengthening the K12 brand. The most consequential advantage of all.

Historically, education functioned as a localized monopoly, with students attending schools based on residential districting. Prior to the pandemic, virtual schooling itself was not widely understood, and awareness of specific options was limited. Virtual education is now broadly recognized by families as a category. The market has now become crowded. In a state like Ohio, there are fifteen statewide virtual charter schools and over one hundred district-operated online programs.

The question is no longer whether virtual school is an option, but which one to choose.

In the past, that decision was heavily influenced by marketing from an education management organization or a virtual school. Schools competed for visibility through search engine optimization, attempting to appear at the top of Google results to get a user to click on a link and drive traffic to a landing page. This was a scalable model to some degree, where upfront investment in SEO could generate incremental enrollments at a low marginal cost. Paid search, bidding on Google keywords, supplemented SEO from an Internet advertising perspective. Yes, the company also invested in direct mail, print media and television commercials, but the presumption is that most of their ad spend was on the Internet.

Given Stride’s scale and the prominence of the K12 brand, ad spend only represented about 4-6% of the company’s total revenue. Presumably a Stride customer would have to pay more in advertising as a % of revenue if they didn’t have access to the K12 brand.

How much would a school need to invest in ad spend today given that AI and LLMs have fundamentally changed how consumer brands are surfaced and become visible on the web?

Discovery is no longer about a family navigating a list of links. It’s now about receiving a synthesized answer. When a parent enters a search today, Google AI Overviews or ChatGPT provides a direct response, and that response privileges sources with established authority across the web.

This shift disproportionately benefits brands with a long history of distributed presence across key websites. Stride’s K12 brand has been cited, reviewed, and referenced across the ecosystems that large language models rely on, including YouTube, Reddit, Wikipedia, LinkedIn, etc… That accumulated digital footprint becomes a structural advantage in an AI-driven discovery environment, increasing the likelihood that K12-affiliated schools are surfaced as trusted recommendations.

This dynamic has important implications for customer retention of virtual schools. For a school board, leaving Stride in the past would have resulted in lower enrollment and reduced marketing efficiency. Now though, disaffiliating from the K12 brand could reduce a school’s visibility within AI-generated answers. It’s uncertain how a school can rebuild that authority on the Internet. Because these AI systems operate as black boxes, there is no clear pathway for a new or unaffiliated brand to replicate K12’s positioning.

In that context, the K12 brand evolves from a marketing tool into a defensive moat. In the pre-AI world, discovery was driven by marketing execution. In the AI-driven world, it is driven by accumulated authority. Stride has built that authority over twenty years. It may prove to be the company’s most durable competitive advantage.

School boards will presumably recognize this and, as a consequence, will be more reluctant to sever their affiliation with the K12 brand.

To test this, I asked Google AI Overviews in Incognito mode (to avoid any personalization bias) to identify the best virtual schools in Ohio. The first result was Stride’s Ohio Virtual Academy (OVA). The output specifically cited K12’s curriculum as a hallmark of the school’s quality, describing it as one of the largest and most established in the country.

I ran the same experiment on ChatGPT and received a nearly identical result.

The transition to a non-K12 brand could potentially cause a school to disappear from the synthesized answers outputted in an LLM-driven world.

In the AI era, the K12 brand may actually be Stride’s strongest defensive moat.

Curriculum Commoditization: A Risk That Requires Investment

Generative AI introduces a clear strategic risk for Stride by collapsing the cost of educational content creation. What was once capital intensive can now be replicated far more cheaply.

Operating a full K–12 virtual school required a massive upfront catalog. 150+ semester-length courses, excluding electives. At an average development cost of $100K per course, building a proprietary library implies a $15M+ investment. That’s just for the initial roll-out and doesn’t include ongoing maintenance. This sort of prohibitive cost forced virtual schools to rely on third-party providers like Stride. It also limited the number of education management organizations in the market because of the massive upfront fixed cost investment required.

Ten years ago, while working at Pearson, the idea that an AI tool could recreate K12’s curriculum on demand would have sounded like science fiction. Today it’s a legitimate question investors should be asking: when will a large language model be capable of generating a course of comparable quality to K12’s existing catalog without any human intervention in the build process—in 2027? 2028?

If and when that capability arrives, a credible substitution threat exists. If schools can develop their own content, the value proposition underpinning the Stride partnership erodes.

To be clear, Stride provides far more than curriculum. It delivers a bundled suite of services including marketing, enrollment operations, customer support, software, compliance, teachers (in some cases) and administration. Also deficit protection (reducing fees to ensure that a school has a fund balance). But curriculum has always been the core differentiator and the source of brand identity. Any education management organization can operate a school, with varying degrees of success on marketing or compliance. Many schools operate independently. Only Stride can provide the K12 curriculum and its teaching model. As content becomes commoditized, bargaining power shifts toward the customer.

Here’s a good case study to provide some context.

The Ohio Virtual Academy (OVA) serves as one of Stride’s largest customers. OVA serves roughly 16,000 students. Per the school’s annual report, it paid Stride $31.6M for web-based software and curriculum for academic year 2024/2025. This represents nearly 40% of the total fee OVA pays to Stride and 26% of the school’s total revenue. This clearly includes both software and curriculum. The data doesn’t provide the split between the two, so for the purposes of this exercise let’s conservatively assume that digital curriculum is half of the $31.6M, or $15M.

If OVA could hypothetically develop its own curriculum at minimal cost using AI, consider how it might redeploy $15M:

• Reduce the student-teacher ratio (at $100K per teacher, $15M equals 150 additional full-time educators)

• Create micro-campuses for in-person enrichment

• Subsidize paid internships

• Fund dual-enrollment college scholarships

• Sponsor annual international travel programs

Would a school offering those enrichments be more attractive to parents than one running Stride’s K12 curriculum? Possibly. The more important question is what happens at contract renewal if OVA arrives with a credible AI-generated curriculum and the ability to threaten full in-sourcing. The leverage dynamics shift materially.

Commonwealth Charter Academy (CCA) in Pennsylvania offers a useful precedent. In the 2010s, CCA developed its own proprietary curriculum and severed its relationship with Connections Education. In 2025, CCA generated over $665M in revenue with a $67M operating surplus. Most schools could not replicate that path given funding realities. Could OVA? In 2025, OVA reported a $5M fund balance, meaning it might have the financial cushion to survive a separation from Stride. But as we saw with Insight of PA, a new governor can change funding to virtual schools and just like that a fund balance can turn into a deficit.

The complicating factor is that families likely chose OVA precisely because of K12’s curriculum and pedagogy. Even if OVA in this hypothetical example claimed its proprietary curriculum replacement was superior with better outcomes, how many students would it be willing to lose in the transition? That student retention risk constrains the board’s leverage.

AI Is Different From Open Educational Resources

Some readers will point out that the promise of free curriculum isn’t exactly new. We’ve seen this before in the form of Open Educational Resources (OER). Free content like EngageNY never truly displaced proprietary curriculum. Why would AI be different?

This is a fair point.

OER are teaching and learning materials. Textbooks, lesson plans, videos, full courses, learning modules, etc… They reside in the public domain or are released under an open license (like Creative Commons). This allows anyone to use them without a cost. The analogy to open source software is apt. And like open source software, OER never achieved the displacement that proponents predicted.

One of the reasons why related to the total cost of ownership. Adopting OER required labor-intensive work: mapping free lessons to state-specific standards, training teachers, providing technical support. The savings on content were offset by the cost of implementation.

It’s unclear what the cost of implementation with AI might be. Again, everything I am describing is theoretical.

AI presumably can automate the mapping, the standards alignment, the adaptation to local requirements, etc… The labor costs that made OER impractical possibly apply or possibly don’t in the same way to AI-generated curriculum.

The structural barrier that protected K12’s curriculum was not merely the upfront investment, it was the ongoing operational complexity of deploying it.

The only way we’ll really understand the risk factor associated with deploying an AI generated curriculum is a school actually doing it. To date, I’m unaware of such an experiment.

To Maintain Its Moat, Stride Must Prove Its Curriculum Is Superior to AI-Generated Alternatives

Historically, Stride’s revenue was protected by the sheer capital intensity of curriculum development. To maintain its current pricing power, Stride must demonstrate that its curriculum is fundamentally superior to good-enough AI-generated alternatives.

Stride is unmatched in K–12 edtech at scale. We are talking about a company that may generate over $2.5B in its current fiscal year. That scale gives it advantages in building AI-driven curriculum that smaller competitors cannot replicate.

That scale will be helpful, because as the cost of generating content approaches zero, the investment required to remain differentiated increases.

Stride must build a next-generation adaptive curriculum offering that is personalized, continuously improving, and deeply integrated into the learning experience. Rather than harvesting margin from lower content creation costs, the company may need to sustain or even accelerate curriculum investment to preserve its positioning.

The notion that companies must aggressively reinvest in their products to harness AI is a theme playing out across software and data sectors. Factset, DocuSign, Zoom, and Microsoft have all explicitly tied margin pressure to AI investment. I wouldn’t be surprised if Stride management announced that next year would serve as an investment year in margin for research and development for the next generation of an AI powered adaptive curriculum. To be clear, management has not communicated their intention to do so in any way.

One potential mitigating factor: Stride capitalizes its curriculum development and amortizes it over three to five years. A surge in development spending today might not translate directly into a near-term margin blow. The cost could be smoothed over multiple years.

Management has not yet had this conversation with investors about investing for the future. This year’s focus has been on LMS stability. The risk is that the company arrives at an epiphany about investing only after a major customer announces a departure.

To the extent that management referenced investment in AI, this is what the CEO said in the 4Q25 conference call:

And of course, everybody is talking about AI. We are proceeding with our cautious but ambitious approach to enable the use of AI in our programs in a responsible and impactful manner. I said a couple of years ago that we are not going to play into the hype around AI but rather focus on foundational areas and technologies that we can leverage for better customer outcomes and experiences. And we are continuing down that path both in partnership with other providers as well as through proprietary.

The lack of specificity on investments regarding AI may have been appropriate six months ago. But after the SaaSpocalypse, investors need better answers regarding the company’s product roadmap.

Curriculum development by Stride would likely focus on adaptive curriculum. Stride specifically referenced adaptive learning technologies in its 10-K

Unlike digital courseware, adaptive curriculum could use AI to identify a student’s specific struggle areas in real time. Instead of a linear path from Lesson A to Lesson B, an adaptive system would function as a real-time instructional agent. For example, if a student falters on a concept, the AI identifies the root struggle and branches the student into a custom practice module or triggers a specific teacher intervention.

Stride services over 200,000 students. The company has the financial resources to build proprietary AI-native products and spread that cost across a massive base.

This is how Stride can strengthen its moat.

Investor Sentiment: Is the Market Pricing in an AI Risk?

Because Stride has been navigating self-inflicted operational challenges, the long-term strategic implications of AI may not yet have registered with most investors. The past six months have been dominated by the fallout from the Canvas and PowerSchool implementations.

Yet short interest in Stride has increased significantly over that period, climbing from 4.4M shares in late 2025 to nearly 6.6M shares as of March. Short interest as a percentage of the float now stands at 22%, making Stride the 230th most-shorted stock in the market, just ahead of Pinterest.

What’s driving the short position?

Is it a continuation of the operational fallout thesis, now largely resolved?

Or is the market beginning to price in a structural risk, the same AI-driven disruption thesis emerging across software and data companies?

Or maybe it’s something else altogether. Stride has been a target for shorts for years. For-profit education stocks always receive scrutiny.

Conclusion: AI Risk Exists, But It Can Be Managed

In my opinion, the core investment thesis for Stride remains intact in the age of AI.

Valuation multiples may expand as investors have comfort that the company has stabilized operations following the disastrous LMS migration to Canvas. Conceivably LRN can once again warrant a 10x EV/EBITDA multiple. Even achieving the historical median multiple of 7x would result in stock price appreciation. It would be helpful if the company posts a good Q3 print in two weeks.

I find it difficult to envision widespread customer attrition in the near term. School boards that endured the fall 2025 disruption are unlikely to subject families to another transition so soon. The switching costs—operational, financial, and reputational—remain high.

So I feel pretty comfortable right now owning Stride stock even with each new headline about AI.

AI likely does not serve as an existential risk for Stride in the same way that it has affected data and services companies.

It does reduce the scarcity value of curriculum and introduces a credible long-term substitution risk if Stride’s product remains static.

Stride likely has time on its side to invest and maintain a product lead over alternatives.

Stride operates under multi-year contracts with school boards. Switching providers is operationally complex. Even if a school believes it can replicate elements of Stride’s offering using AI, the transition would require time, coordination, and a willingness to accept execution risk. That creates a natural buffer.

Pricing pressure, not customer attrition, becomes the first signal of disruption.

The greatest risk is not AI itself, but underinvestment in response to it. If Stride attempts to expand margins by reducing curriculum and software investment, it risks eroding its differentiation at precisely the moment when customers gain alternatives.

Stride has an opportunity to out-invest the competition and ensure that the K12 experience remains the gold standard.

Disclosure: I am long LRN at the time of this writing.

Disclaimer: This is for entertainment purposes only. Do your own research. This analysis is my own and is not intended as investment advice.

• Google AI Overviews are breaking the link between search queries and website traffic. Visibility now depends on whether content exists inside the sources LLMs cite.

• Despite deteriorating economics driven by Google AI Overviews, companies have been able to adjust their business models and tactics to survive and thrive in a challenged environment. In the piece we review the examples of three companies who have adapted successfully.

• Case 1: Higher Ed. University of Phoenix fixed lack of visibility in AI Overviews by moving outcomes content to YouTube.

• Case 2: Publishers. People Inc has replaced lost traffic with licensing, platform distribution, and commerce.

• Case 3: Lead gen companies. NerdWallet has shifted its model toward paid acquisition and owned ecosystems.

• Entities need to adjust their marketing, the destination is no longer the website. Leaders now need to managing their total digital footprint.

In the 1980s, Ralph Macchio played a young man on a journey. He sought out a seasoned mentor, learned a difficult craft, faced a seemingly superior opponent in a high-stakes climax, and, against all odds won.

I’m not talking about The Karate Kid.

Two years after Daniel LaRusso did the crane kick, Macchio starred in a much less well-known film called Crossroads. In it, he plays a Juilliard student trying to master the blues. The movie follows the exact same narrative beats as The Karate Kid, but it ends with a supernatural guitar duel against a demonic opponent played by rock legend Steve Vai.

Same actor. Similar setup. Completely different story.

There is a parallel here to how organizations develop their brand in the AI era. We are using the same actors (C-Suite / marketing teams) and the same setup (search engines), but we are in a completely different story.

The New Normal is that Discovery has moved outside of an institution’s control. Large Language Models (LLMs) are affecting user discovery not by sending users to a destination website but by answering questions directly by pulling information from “trusted” authority sites like Reddit, LinkedIn, Wikipedia, and YouTube.

The Thesis: Marketing as Portfolio Management

This New Normal requires more than just fixing a website. It potentially requires rethinking an institution’s business model and the profession of marketing.

My thesis: Marketers are Portfolio Managers of External Content. The profession isn’t just about driving traffic to a website.

This was validated this past week when Phoenix Education Partners, owner of the University of Phoenix, announced their most recent financial results. Management suggested that a change in the Google AI Overview algorithm had directly affected their enrollment pipeline.

As anyone doing a basic web search knows, Google now provides a synthesized answer instead of a list of links. The problem Phoenix saw was that because they lacked content on YouTube, they were invisible in those synthesized answers. Or perhaps differently, AI was filling the void with messages that didn't align with their brand.

Their solution? They created content on YouTube. According to their conference call, that pivot solved the problem.

Other companies have complained about the deleterious effects of Google AI, but this is the first circumstance that I’m aware of where a management team addressed the specific type of content being harvested.

Right now, leadership of companies or institutions of higher education need to obsess over visibility inside the answer layer of AI. A successful marketing function in 2026 will figure out:

1. How to show up in the AI Overview answer

2. How to ensure that the LLM cites its data as the “truth”

This article is intended for stakeholders of laggard institutions. Institutions that aren’t making meaningful changes to their business model or how they market. Readers need to understand the severity of the problems that Google AI Overviews has caused.

Some companies (e.g., Chegg) are on the road to extinction.

Others though are evolving and growing with a model shift.

There are real answers to the existential problem that Google AI Overviews creates.

I have identified three distinct types of failures that I’ve seen from Google AI Overviews, but also three publicly traded companies that have made adjustments to their business model and tactics employed.

The Publisher Failure: Content is Used, But Traffic is Lost

In 2022 media entities announced a new Karate Kid movie would be developed.

If a user wanted to learn more about the movie they would have typed in a web search most likely on Google. Links to various websites related to the movie would have emerged. Perhaps the first link would have been to Wikipedia, the second might have been to Reddit, and the third might have been to People.com.

People.com relied on users landing on the website page and monetizing users via ads or subscriptions.

The rise of Google AI Overview broke this model.

If you want to read about the next Karate Kid movie, Google AI Overview will provide you all the info you’d ever want.

This hasn’t been a good development for publishers. For example, IAC’s People Inc saw a 50% drop in Google Search referrals over the past two years. By 2025, Google search accounted for less than a third of the company's traffic, down from 60% in 2021.

The Wall Street Journal reported on this dynamic this back in June 2025. This publisher issue is far from new. In 2026, Define Media Group suggested that organic searches for publishers had declined by 42%.

People management didn’t try to fix their search engine optimization. Instead, they built a business model that assumes Google traffic isn't coming back.

1. AI licensing. They signed deals with OpenAI, Microsoft, and Meta to get paid for the content the AI uses, creating a revenue stream that does not depend on website visits.

2. Non-Google distribution. They moved where other users are—TikTok, Apple News, and YouTube—nearly doubling off-platform views in two years.

3. Commerce. They pivoted toward monetization through purchases, reducing their reliance on raw traffic volume.

By the end of 2025, digital revenue rose 14% despite the traffic collapse. They stopped being a destination and became a licensed content supplier.

The Lead Generation Failure: No Click = No Funnel

Publishers want users to land on their pages and consume content.

Lead generation companies, like NerdWallet or Bankrate or Quinstreet, create content as a catalyst for transactions. They create high quality guides to inform consumers about product offering and then connect the user to that service. Think of a website where you learn about an offering and then hit “apply now”.

In the old search model, a user typed "best travel credit cards," clicked NerdWallet, and was guided to a bank's offer. Today, Google AI Overview has become the advice layer itself. If the AI tells the user which cards are best, the user may go directly to the bank, bypassing the intermediary.

This obviously has negatively affected the lead gen companies.

Publicly traded NerdWallet (NRDS) saw meaningful business challenges in their most recent Q4 results. Credit card revenue declined YoY by 24%. SMB products revenue declined by 12%.

But they adapted. They re-engineered their business model. NerdWallet revenue for 2025 overall increased by 22%. They did it through:

• Vertical Integration: They moved away from being just an intermediary. By acquiring the mortgage brokerage Next Door Lending, they became a broker, capturing a higher profit margin which provided them a cushion to pay for more expensive traffic.

• Performance Marketing: They increased their spend on sponsored links by 40%. They stopped paying for broad brand awareness, like Super Bowl ads, and started buying the top spots on Google to guarantee a click.

• Direct Relationships: They use paid traffic to drive users into their own app. Once a user links a bank account, NerdWallet can reach them via notifications, bypassing Google entirely.

NerdWallet is no longer just a lead generator; they are building a private ecosystem to protect themselves from the decay of organic search.

The Institution Failure: Omission from the Synthesis

In the first two cases, companies had to make structural changes to their business models. People Inc. moved toward licensing and platform-native distribution to survive the loss of site traffic. NerdWallet pivoted to high-margin verticals and direct app engagement to protect their revenue from the loss of the organic search click.

The case of the University of Phoenix represents a different type of disruption. A visibility failure. Granted, the financial impact from Google AI Overviews was modest compared to the other cases. Nonetheless, it revealed the new mechanics of discovery.

During their Q2 2026 earnings call, management noted that an “algorithm shift in AI search on Google” changed how prospective students navigated their search. Google’s AI was prioritizing YouTube as a primary source of information.

Phoenix has always possessed outcomes-based content proving that their programs lead to career advancement, but that data didn’t reside on YouTube. Google AI didn't pull this data into the synthesized AI answer. Phoenix’s value proposition was omitted.

Their solution included migrating outcomes-based data onto YouTube.

While Phoenix’s immediate problem has been resolved, it demonstrates the threat of AI affecting their enrollment funnel in the future. Content on third parties clearly will play a meaningful role in their financial results on a go forward basis. Historically Phoenix has had one of the best marketing capabilities around. So they likely will figure it out on a go forward basis. The same cannot be said of its higher ed peers.

The End of the Destination Website and the Rise of Managing the Digital Footprint

While Google still has 90% market share of search, its expected that this will whittle down in size with other LLMs gaining share over the coming years. As specialized LLMs gain traction, we will enter a fragmented era where discovery happens across multiple proprietary search mechanisms. Each with their own idiosyncratic methods and sources.

Like with publishers and lead generators, academic institutions may see an existential crisis.

The traditional model of buying keywords to drive traffic to a homepage where a student can apply is done.

Game over man, game over.

This new reality doesn’t have to end poorly though for institutions with forward thinking leaders. Institutions have options. As the cases of People, NerdWallet, and the University of Phoenix show, there are ways to adjust.

One place for leaders to start is rethinking the marketing function. Marketing is now about managing a Portfolio of External Content.

To maintain visibility, institutions will need to have consistent messaging across YouTube, Reddit, Wikipedia, LinkedIn, and other authority sources.

Higher ed institutions must now manage the data supply that feeds these models. The organizations that survive this shift will be those that realize their reputation isn’t what they’ve written in their marketing collateral. It’s the aggregate/synthesis of what the most trusted sources on the web say about them.

Historically higher ed has not adjusted with speed to market conditions. That explains the success of the for-profits. Business owners adjusted quickly to market conditions. Nonprofits responded over the course of years. The for-profits embraced digital delivery for working adults. The nonprofits eventually realized the financial opportunities to service this same student population.

Given the speed of AI though, slow adaptation may mean closure.

There is a final chapter to the Ralph Macchio analogy that I provided in the beginning of this article.

For decades, Macchio was a nostalgia act. He wasn’t part of the current cultural conversation.

Then came Cobra Kai.

Macchio’s career resurgence happened because he embraced a new platform (YouTube). He took advantage of the modern fragmented streaming era. He stopped waiting for Hollywood to find him and instead went where the new audience was already living.

For laggard institutions, the lesson is the same. They cannot wait to be found on Google anymore.

In the age of the AI synthesis, doing nothing is a death sentence.

• While AI may be driving sector re-ratings, valuation in edtech still holds on a relative basis

• Across edtech, relative valuation is anchored in fundamentals - primarily revenue growth, not irrational AI narratives

• Coursera (COUR) at ~3x EV/EBITDA looks distressed, but on a relative basis with current consensus estimates, it’s trading exactly in-line with its peers. And if/when the merger with Udemy (UDMY) goes through, the stock could see valuation multiple expansion if the regression model holds

• Duolingo (DUOL) stock may have declined by 78% from peak, but on a relative valuation basis it’s trading where it should relative to peers

• Although Stride (LRN) lacks a clean comp set, when anchored to postsecondary peers, it appears as undervalued with clear multiple expansion potential

• Postsecondary stocks have split into three distinct buckets:
  vocational (premium valuation), healthcare/ground campus (mid-tier valuation), and online (discounted valuation)

• Bottom line: AI hasn’t changed the rules of valuation—fundamentals still determine where stocks trade

Has AI Changed How Publicly Traded Edtech Companies Are Valued?

HBO commentator Larry Merchant famously said during the Mike Tyson - Buster Douglas fight, Douglas is asking of Tyson some questions he hasn’t been asked before

Douglas, at that point, was doing the unthinkable—breaking a model everyone assumed was untouchable.

We’re seeing something similar in the world of edtech given AI disruption.

We assume a company will exist because it already has. That’s what’s called normalcy bias. That’s when people believe things will continue functioning as they always have and ignore impending disasters. Investors already played that game with Chegg and got burned, with the stock losing close to 99% of its value. That business collapsed over the span of two years given AI substitution of its core product.

In edtech, basic questions are no longer theoretical:

• How will people actually learn?

• What is the value of a credential or degree in a world where content is free?

• Why subscribe to a content platform when AI can generate answers on demand?

• Which job categories are truly insulated from AI substitution?

For investors, the more immediate question is simpler:

How do you even value these businesses?

In this post Q1 write-up, I’m focusing on whether edtech relative valuation has merit.

Relative valuation determines how a company is priced in relation to its peers, typically using multiples such as EV/EBITDA. It is, in practice, how the public markets price assets.

With AI driving volatility across digital and edtech names in recent months, it’s reasonable to ask whether valuations still make sense.

Take Coursera as an example. The stock trades at roughly ~3x EV/NTM EBITDA, a level that typically implies structural decline. Yet the business is still growing, and to date has not seen revenue materially impacted by AI.

So what explains Coursera’s valuation multiple?

Why is it trading below the valuation of postsecondary companies?

Is the market even pricing companies relative to peers?

And if so, what drives differences in valuation across the group?

Approach and Findings

To answer these questions, I constructed a series of relative valuation frameworks across edtech and education companies.

For each company, I:

• Identified a relevant peer group

• Mapped valuation (EV/EBITDA) against measures of financial performance

• Evaluated how tightly valuation clusters around those performance metrics

The objective:
Does financial performance still explain valuation dispersion across stocks?

My conclusion:

AI has clearly influenced sector valuation multiples.

But at the company level the underlying valuation framework remains surprisingly intact.

Across the names analyzed, differences in valuation are still largely explained by revenue growth. For Coursera, revenue and profitability.

AI may be shifting the level of valuations, but it is not redefining the rules by which companies are valued relative to one another.

In this piece I focus on the following companies:

1. Coursera: COUR

2. Duolingo: DUOL

3. Stride: LRN

4. Postsecondary stocks: STRA; PRDO; UTI; APEI, CVSA; LOPE; LINC; PXED

As with any relative valuation exercise, results are inherently sensitive to peer group construction. Including or excluding companies can impact the explanatory power of the regression.

The peer sets are designed to reflect similar business models, comparable scaling dynamics, and analogous end markets. The goal here is not precision though. We are working with inherently imperfect comparables.

To avoid the temptation of creating peer groups for the expressed purpose of supporting the conclusions I want to create, I initially used ChatGPT to generate peer sets and then applied judgment to exclude companies that deviated too far from the target company’s operating model.

Coursera Valuation: Trading in Line with Marketplace and Consumer Stocks

Coursera is best understood as a marketplace business with a subscription revenue model. It connects learners, institutions, and enterprises, facilitating the distribution of content at scale. The underlying economics are consistent with other marketplaces: aggregation of demand and supply; operating leverage as volume grows; network effects.

The stock has fallen by half since last summer. In December management announced a merger with its competitor Udemy. At the time I expected the transaction to act as a catalyst for stock price appreciation. My rationale: increased scale with $115M of identified run-rate cost synergies would meaningfully improve profitability; the acquisition would give the new company at least a year to fix Udemy’s business model. On the stock price performance, I was wrong.

To evaluate Coursera’s valuation, I compare it against a group of consumer and marketplace businesses, including Roblox, DoorDash, Pinterest, HubSpot, Wix, Adobe, monday.com, Klaviyo, and Duolingo.

Yes, these companies operate in different end markets like gaming, food delivery, content, etc… But they share similar characteristics: demand aggregation, user engagement, monetization layers, and operating leverage.

Roblox, for example, may be a game that your kid plays, but it monetizes user-generated content and transactions. DoorDash and Uber are more traditional marketplaces. Wix and HubSpot layer monetization on top of a large base of users.

Since the start of 2025, Coursera’s stock has lagged my constructed marketplace index, the S&P 500, and software benchmarks such as the iShares Expanded Tech-Software ETF.

This underperformance primarily reflects multiple compression. The company consistently exceeded quarterly consensus estimates.

Regarding valuation, in the past investors were forced to value the stock on a EV/Revenue basis because the company didn’t generate profits.

Following the SaaSpocalypse, there has been an investor shift toward profitability, making EV/EBITDA the more relevant valuation metric.

Using this framework, the peer group shows a reasonably strong relationship between EV/EBITDA and revenue growth, with an R² of approximately 0.69.

As a reminder, a R² is the coefficient of determination, it measures how well a regression model can explain variation in data.

I also constructed a regression using an adjusted Rule of 40, which combines revenue growth and profitability into a single metric. The Rule of 40 is adjusted to weight revenue growth more than profitability.

When EV/EBITDA is plotted against the adjusted Rule of 40 of this comp group, the relationship strengthens meaningfully, with an R² of 0.82. Valuation across this group is better explained by the combination of growth and profitability than by growth alone.

In both frameworks, Coursera sits at the lower end of both regressions.

Its valuation aligns with its current growth and margin profile (as a standalone company).

In other words: the stock is trading where the model would suggest it should.

Well, not exactly… the current consensus estimates reflect the standalone business and do not incorporate the potential impact of the Udemy acquisition. $115M in expected cost synergies should improve EBITDA margins. Revenue by definition will improve optically because the combined company will be viewed relative to standalone Coursera revenue. I have no idea on a pro forma basis how the combined company will look, but stock screens don’t provide info to investment analysts on a pro forma basis.

If the combined entity delivers higher revenue growth and improved margins, the implication within this relative valuation construct is that a higher adjusted Rule of 40 should support a higher multiple.

Let’s just hope that Coursera and Udemy shareholders approve this merger in April and get the ball rolling. The stock could potentially move following management providing post-merger revenue and EBITDA guidance.

Duolingo: Trading In-Line with Consumer and Freemium Oriented Peers

Duolingo is a freemium, gamified language-learning app. Yes, it doesn’t have a clean comp set. That makes it difficult to value on a purely relative basis.

That said, the company shares characteristics with a range of publicly traded platforms:

• Global scale with broad user reach

• High-frequency engagement

• A free-to-paid conversion funnel

• Operating leverage

To benchmark valuation, I constructed a peer group that includes Roblox, DoorDash, Pinterest, HubSpot, Wix, Adobe, monday.com, Coursera, and Klaviyo. These companies operate in different verticals but again, they share common economic structures. The businesses are driven by usage, conversion, and retention.

Like Coursera, Duolingo has underperformed relative to its peer group. Since the start of 2025, the stock has lagged a broader consumer/freemium index, the S&P 500, and software benchmarks such as the iShares Expanded Tech-Software ETF. The stock saw meaningful stock price appreciation in mid-2025, but has since given back those gains, reflecting both multiple compression and concerns regarding the durability of its business model in an AI world.

Unlike Coursera, Duolingo’s valuation aligns with the traditional relationship between EV/EBITDA and revenue growth. Not adjusted Rule of 40 like Coursera.

When EV/EBITDA is plotted against revenue growth, the relationship is strong, with an R² of 0.81. In contrast, when valuation is plotted against an adjusted Rule of 40, the fit weakens materially, with R² declining to 0.50.

Within this peer set, Duolingo does not show up at the high end of either growth or valuation. Rather, it sits in the middle of the group, with a multiple that is broadly consistent with its current financial profile. If you take a look at the chart, it actually sits on the regression line. Meaning that at the current level on a relative basis the stock is neither a screaming buy nor a short.

Stride - K-12 Operator Valued Within a Postsecondary Framework

Stride is primarily a K-12 education company whose operating model is driven by virtual school enrollment and funding levels. It is the only pure-play publicly traded K-12-focused operator, with no direct peer set. While Pearson owns a competing asset (the division formerly known as Connections Education, my professional alma mater), its broader mix of businesses makes it an imperfect comp for valuation purposes.

There is no clean comp for this company. To the consternation of its investor base.

Despite being an orphaned stock, my sense is that investors for better or worse anchor the company within the broader peer group of publicly traded postsecondary education operators like Strategic Education, Perdoceo, Universal Technical Institute, and Lincoln Educational Services. Those operators are valued primarily through a relationship between revenue growth and EV/EBITDA multiples.

The postsecondary peer group exhibits a strong relationship between growth and valuation, with an R² of 0.82. Including Stride in the regression reduces the R² modestly to 0.80, suggesting that while it is not a perfect fit, it still aligns reasonably well within the same valuation framework.

Historically, Stride traded at 5-6x EV/EBITDA. During the pandemic, enrollment tailwinds drove stronger revenue growth which in turn increased revenue and profitability. Interestingly, as you can see from the chart below what drove multiple expansion began in earnest in fall of 2023. Perhaps with the realization that the virtual school proposition was a secular growth story rather than a temporary cyclical Pandemic benefit. At its peak over the summer of 2025, the stock traded closer to 10x EV/EBITDA.

More recently, the stock declined by over 50%, driven in part by operational challenges related to a transition to a new learning management system. When the company reported earnings in January, the stock price increased over 20% as investors saw that the tech issues had stabilized.

When viewed through the regression of EV/EBITDA against revenue growth for the broader postsecondary peer group, Stride appears undervalued on a relative basis. Based on the company’s current growth rate, the implied multiple is closer to ~8.5x EV/EBITDA compared to its recent trading levels nearer to ~6x.

If Stride posts a good Q3 print, this valuation framework potentially implies that the valuation multiple will expand up to where the regression suggests it belongs.

Postsecondary: Valuation Anchored to Revenue Growth

In Q1 2026, the postsecondary group materially outperformed the S&P 500, delivering high double-digit returns versus a roughly -7% decline in the broader market. This divergence suggests a sector-specific re-rating, rather than a macro-driven move.

That performance, however, was highly concentrated. A small number of names—most notably Lincoln Educational Services (+69%), American Public Education (+51%), and Universal Technical Institute (+42%)—accounted for the majority of the gains.

The best model to explain the sector’s relative valuation is the relationship between revenue growth and EV/EBITDA. The key drivers—enrollment trends, pricing, and program mix—flow directly into revenue growth, which in turn drives valuation multiples.

Higher-growth institutions command higher EV/EBITDA multiples. Lower-growth operators trade at a discount.

What’s more notable is how tightly these companies cluster around a regression line.

Three distinct groupings emerge:

• Vocational / campus-based operators (UTI, LINC) sit at the high end—both in terms of growth and valuation. These businesses appear to benefit from strong demand and clearer alignment with workforce outcomes. Their underlying professions may also be less susceptible to near-term AI disruption (e.g., skilled trades).

• Nursing and ground-campus exposure (APEI, LOPE, CVSA) cluster in the middle. They combine moderate growth with mid-tier valuation multiples. AI won’t replace nurses.

• Online-heavy models (STRA, PRDO, PXED) screen at the low end, with lower growth and correspondingly lower valuation multiples. These models may face greater competitive pressure and potential substitution risk in certain white-collar pathways.

The divergence in valuation across postsecondary operators becomes even more evident when looking at multiple trends over time.

Vocational operators have seen the most pronounced re-rating. Both LINC and UTI have experienced sustained multiple expansion over the past 12–18 months, with forward EV/EBITDA multiples now at or near cycle highs. This reflects a combination of accelerating growth, improved execution, and investor preference for workforce-aligned education models.

In contrast, nursing and ground-campus institutions—including American Public Education (APEI), Grand Canyon Education (LOPE), and Covista (CVSA)—have exhibited more stable valuation trends. Multiples have generally moved within a narrower range, consistent with moderate but steady growth profiles and less volatility in underlying demand.

The most notable divergence is in online-focused operators. Strategic Education (STRA), Perdoceo (PRDO), and Phoenix Education Partners (PXED) have seen flat to declining multiples over the same period, with valuations remaining closer to cycle lows. These businesses have not benefited from the same multiple expansion, reflecting slower growth and increased skepticism around long-term positioning.

The data highlights the widening spread in valuation across programs. Vocational operators are being re-rated upward, while online models remain compressed, reinforcing a clear barbell in investor preference.

Conclusion

Despite the narrative around AI, the market is still valuing education and edtech companies using familiar frameworks.

At the individual company level, valuation remains grounded in financial performance. Revenue growth and profitability still explain where stocks trade. AI may be shifting the level of valuation multiples, but not the underlying relationship between growth and valuation.

Companies are still being rewarded based on what they actually deliver.

So for operators reading this piece, at least you know that the market isn’t acting totally irrational, and that you do have some say in how your company is trading in a post-AI world.

Disclaimer

This analysis reflects my personal views and is provided for informational purposes only. It should not be construed as investment advice or a recommendation to buy or sell any securities. The information presented is based on publicly available data and my own analysis, and while I believe it to be reliable, I make no representation as to its accuracy or completeness.

Investing in public equities involves risk, including the potential loss of principal. Readers should conduct their own due diligence and consult with a financial advisor before making any investment decisions.

I currently hold positions in Coursera (COUR), Udemy (UDMY), Stride (LRN), Phoenix Education Partners (PXED), and Strategic Education (STRA). My views may change at any time without notice.

This analysis was created with the assistance of AI for editing and visualizations.

• LLMs and AI search are changing how colleges are perceived by surfacing indications of financial difficulty, including solvency concerns.

• Institutional financial challenges, which were once opaque, is now immediately visible at the start of the enrollment process. Parents can now see this financial risk before ever visiting a website or engaging with admissions.

• For financially weak colleges, even small declines in enrollment can trigger operating deleverage that quickly triggers a financial crisis. This creates a feedback loop: financial concerns → fewer students → worsening finances → greater concern.

• Hampshire College illustrates the risk — its financial challenges and going concern warnings are now surfaced directly in AI-generated answers. Will already challenged enrollments deteriorate with increased transparency?

• Institutions need to act now: understand how they appear in AI-driven search, strengthen their digital footprint and third-party validation, and develop a clear plan to address both perception and underlying financial risk in a fully transparent environment.

Higher education leaders are underestimating how quickly AI will reshape institutional durability—and in some cases, accelerate institutional failure.

Consider Hampshire College.

Hampshire College is a small liberal arts college with a distinctive, non-traditional academic model. Founded in 1970, the institution emphasizes self-directed learning, individualized concentrations, and uses narrative evaluations in place of traditional grades.

When I was at Amherst College, part of the Five College Consortium, Hampshire students clearly stood out. They were unconventional, willing to take a different path in life. The school lacked the more structured “corporate” feel of traditional elite liberal arts colleges. That was part of its appeal.

It has also faced persistent financial challenges.

In 2019, Hampshire’s auditors issued a going concern opinion, indicating significant doubt about the institution’s ability to continue operating within a one-year period. Leadership announced that the college might not admit a full incoming class and began exploring strategic alternatives, including a potential merger.

The situation escalated quickly. Students protested, occupied administrative offices, and drew national attention to the school’s future. Students occupied the president’s office for 75 days. The school drew national attention regarding the institution’s future. Filmmakers made a movie about it.

Hampshire ultimately avoided closure. It downsized operations, restructured, and raised funds to stabilize itself, ultimately surviving the pandemic as a smaller institution.

In 2026, its auditors once again issued a going concern opinion. Its accreditation is now at risk. The institution appears to be operating at a loss.

As you can see from the chart below, Hampshire’s cash and investments have continued to decline since 2020. It has a lower cash balance than it did during its 2019 crisis.

Note that its endowment draw as a percentage of operating expenses exceeded the typical industry norm of ~5%. This was not a sustainable path for them over the long-haul.

Now introduce AI into the equation.

I asked ChatGPT a simple question.

My daughter may apply to Hampshire College, what are your thoughts?

The response did not just describe the academic model or student experience. It surfaced concerns about financial instability.

When I asked the same question to Google, AI Overviews produced a more balanced response—acknowledging risk, but also emphasizing the institution’s unique model.

ChatGPT suggested that there were risks, whereas Google AI Overviews suggested that the institution’s unique model was worth those risks. The issue is not which system is right. The issue is that this information is now surfaced, and surfaced immediately.

This is a remarkable shift in public transparency about an institution’s financials.

Historically, while this type of information existed it wasn’t front and center. A parent might encounter it later in a process, perhaps after visiting campus or speaking with an advisor. Or maybe not at all.

Now, it appears at the very beginning of the enrollment process.

The narrative of an institution is no longer controlled by the school. It is synthesized from across the Internet from the institution’s digital footprint.

AI Will Compress the Enrollment Funnel

My thesis, the point of this piece: AI will accelerate the decline of financially challenged schools by compressing the top of the enrollment funnel.

Industry practitioners typically think about the enrollment cycle as a funnel.

Prospects → Inquiries → Applicants → Admits → Deposits → Enrollments

Everything starts at the top. The funnel narrows with each step of the process. Perhaps 50K prospects become 1K enrollments.

In the past, a student or parent would begin with a Google search, click through multiple websites, and gradually form a view. If a school could get a click, it could tell its story.

AI has broken this model.

Instead of browsing on the web, a user asks a question and receives a synthesized answer. A small number of institutions are recommended. Others are not mentioned at all.

For those that are mentioned, framing matters. A school may appear alongside commentary about outcomes, reputation, or financial instability.

In effect, LLMs are becoming de facto college counselors.

This has a direct effect on the funnel.

If fewer institutions are surfaced, fewer enter the consideration set. If an institution is excluded or included with negative framing, the number of prospects entering the funnel declines.

Fewer prospects → fewer inquiries → fewer applications → fewer enrollments

Discovery Problems Compound Financial Difficulties

For institutions that are already fragile, even small declines at the top of the funnel can have outsized consequences.

Higher education has a high fixed-cost structure. Faculty, facilities, and administration do not scale down easily with enrollment. When revenue declines, costs remain largely intact.

This is operating deleverage. The loss of even a small number of students can materially increase financial risk.

In the case of Hampshire College, you can see that revenue declined from roughly $60M earlier in the decade to roughly $30-35M.

Leadership I’m sure did its best to cut costs, but with enrollment declines and the inability to cut fixed costs the cost per enrollment increased.

AI/LLM Risk Different Than Other Past “Existential Risks”

Yes, colleges have always had varying levels of financial strength.

Yes, institutions have survived significant challenges in the past. Like the recent pandemic.

What is new is that this sort of information is showing up front and center during the top of the funnel.

The income statement and balance sheets of financially distressed institutions have always been opaque to prospective students. Families are not in the habit of reviewing financial statements, reading accreditation reports, or trade journals that describe the issue.

AI changes that.

It surfaces institutional reputation, outcomes, and financial signals at the very start of a process. Before a school has an opportunity to engage the family.

This creates a reinforcing dynamic.

If concerns about financial stability are introduced at the prospect stage, some families may never move forward. They may not inquire, apply, or visit campus.

The funnel narrows before the institution can tell its story.

And because fewer students enter the funnel, enrollment declines.

For a financially fragile institution, that decline can worsen the underlying financial position—further reinforcing the original concern.

How to Think About Institutional Risk in an AI-Driven Market

AI now introduces a new variable into what already has been a financially challenged sector.

Two factors now interact:

• How financially resilient the institution is

• How visible and how it is described in AI-driven discovery

A school that is financially strong but not visible has a marketing problem. That can be fixed through a focused marketing effort.

A school that is visible but financially weak has a structural problem. The school has time to consider merger options, launch new revenue generating initiatives, or cut expenses.

A school that is both weak and not surfaced has an existential crisis. Not a lot of time to make changes.

What Institutions Need To Do

I hope that Hampshire College successfully navigates this period. They can fundraise their way out of this. They can seek a merger. But they need to do something immediately.

Institutions that are several years away from this sort situation have the flexibility to start acting now in a structured and thoughtful way.

They need to:

• Diagnose how they appear in AI-driven queries

• Strengthen external validation across third-party sources (outcomes, employer relationships, alumni narratives, media coverage)

• Clarify positioning so the institution is differentiated in AI outputs

• Communicate financial strategy clearly and credibly to stakeholders

LLMs will not determine which colleges succeed.

But they will make it painfully obvious which institutions were already at risk.

And in doing so, they may accelerate outcomes that would have otherwise taken years to unfold.

Don’t Discount How Bad of a Situation This Could Be

AI is arriving at a time when the sector is already under pressure.

The enrollment demographic cliff—a projected decline in high school graduates beginning around 2025–2026—will intensify competition for students.

The Western Interstate Commission for Higher Education projects a ~10% decline in high school graduates by 2042 relative to 2022 levels.

Unfortunately, the next decade or so will see many schools close due to financial challenges. The Federal Reserve of Philadelphia estimated that, in a worst-case scenario, up to 80 institutions could close per year.

This dynamic in itself intensifies competition for students.

Fewer prospective students means fewer enrollments. For tuition-dependent institutions, that translates into lower revenue.

Hopefully those reading this are aware as to how quickly AI can disrupt demand-driven businesses.

Consider Chegg, a publicly traded provider of “homework help”. As LLM-based tools began to replicate its core offering, the company experienced a rapid deterioration in its subscriber base. Billions of dollars in market value were erased over two years. The company’s long-term viability is in doubt.

When AI changes how users access information, demand can shift quickly. And abruptly.

Conclusion

LLMs will make it immediately visible which institutions are at risk.

And by surfacing that risk at the very start of the enrollment process, they may accelerate outcomes that would have otherwise taken years to unfold.

Institutions with weak financial profiles cannot wait.

They need to understand how they are perceived, how they are surfaced, and how much time they have to respond.

Because once these dynamics become visible at scale, recovery becomes significantly more difficult.

Disclosure: This piece reflects my analysis; AI tools were used for editing support and visual generation.

• AI is changing how students choose universities. Prospective students are provided a small set of recommended institutions in a single synthesized answer by LLMs. Marketing is no longer about striving to appear through SEO tactics at the top of Google search results.

• A University website is now just one input among many for a LLM. Institutional visibility is now a function of its total digital footprint across the broader web. Landing pages just aren’t enough.

• LLM inclusion requires/mandates an institution building reputation from external sources on the web. Alumni outcomes, employer relationships, faculty visibility, and public discourse can assist with LLM inclusion.

• Universities must change how they operate, not just how they market. Inclusion in LLM outputs is no longer a marketing function; it is an enterprise-wide operational mandate. Employees across function need to adopt tactics to increase visibility across the internet or risk not being included at all in LLM outputs.

• This piece outlines six concrete actions universities can take to be included in AI-generated answers.

Hopefully at this point in the AI era administrators understand that the adoption of large language models (LLMs) will significantly change how higher education institutions market themselves. If this is in dispute, well, that’s a bigger problem than the one that I’m addressing in this piece.

Currently Google has close to 90% search market share. Some have estimated that LLMs will reach 50% of search market share by 2030. Google AI Overview now appears in 42.5% of all Google searches.

We are living through in real time a structural change in how information is delivered and consumed. It happened so fast that little literature exists on best practices to market to prospective learners in this new reality.

For the past twenty years, institutions optimized their website to rank highly in search results. The premise was to receive free leads and convert that traffic into inquiries and applications. This practice, known as search engine optimization (SEO), required a fixed cost investment in the form of internal marketing staff and consultants. They optimize a website that results in a near-zero marginal cost for each incremental click on a Google provided link. Marketers supplement SEO with paid search, which includes sponsored links and bidding on Google keywords.

AI completely disrupts this customer acquisition model. Not just a little bit. A lot.

But you, dear reader, you already know about this from your day to day search queries on Google. Its hard to miss.

The question is, in a world where AI systems synthesize information and present a reduced set of answers, how do institutions compete for visibility within a LLM conversation?

I’ve published on Substack and LinkedIn on customer acquisition and biases that exist with LLMs (particularly against for-profit institutions). Those thought pieces catalyzed conversations with leadership teams and marketing professionals. I’ve talked with them about the transition from SEO to generative engine optimization (GEO). Or as some call it LLM optimization (LLLMO). Or answer engine optimization (AEO). You know it’s a new space when there isn’t one universal name for a concept.

Takeaways from those conversations:

- Anything I or anyone puts in print will be stale soon after publication

- Folks are not appreciating the level of change required in the new normal.

The reason for this piece though is that everyone I talked with has thought of needed change as incremental.

My take is that the rise of AI and LLMs requires institutional transformation.

The Shift from Website-Centric Discovery to Synthesized Evaluation

Historically, students searched, clicked links, filled out forms, compared options, and submitted applications. Institutions competed for ranking high on a Google search, generating traffic, and converting that traffic into applicants.

Applicants used sources that provided external validation. Entities like US News and World Report. Back when I was in high school these rankings used to be a big deal, a source of conversation and ribbing with my classmates.

I don’t know to what extent people care about an institutions ranking on US News and World Report. As of 2022 people apparently still cared, at least according to the New York Times.

In a world where applicants receive synthesized answers, university and US News and World Report websites aren’t a destination but rather are inputs for synthesized information. Listen to this podcast from a US News and World report employee on his perspective of how it still matters, it’s an interesting take.

He’s right in that it matters to the extent that it’s an external source of validation and thus matters in AI land. But there are others as well that matter. It includes commentary on Reddit / LinkedIn. LLMs use a triangulation logic to determine institutional credibility. A claim made on a .edu domain carries less weight than a claim mirrored across a high-authority ecosystem.

To the LLM, truth is a consensus derived from disparate, high-authority external data points rather than a single primary source.

The Website Is No Longer the Center of the Marketing Strategy

In my conversations, I was told that institutions at minimum need do two things to help with LLM picking up university content. Institutions should:

1) Continue to invest on SEO optimization. Institutions need to ensure that their websites are properly indexed and discoverable. Google rankings still matter, because highly ranked institutions are more likely to be considered as inputs into LLM-generated answers.

2) Optimize the institution website such that it can easily be ingested into the LLMs. This includes replacing marketing language with clear, structured, and answerable content that models can extract. Content should be organized around real user questions, in FAQ-style formats. Also, content needs to be frequently updated.

There’s a whole science about optimizing websites for AI consumption. Lots of tips and tricks.

This is all great info.

The problem though is that these commentaries reflect a continuation of website-centric thinking in a system that is no longer centered on the university website.

External Validation – The New Focal Area for Higher Ed Marketing

If an institution focuses only on its website, it is addressing only part of the new normal.

AI systems rely on signals that exist outside the institution’s control:

• alumni outcomes

• accreditation

• databases of higher ed comparing institutions

• employer relationships

• faculty visibility

• leadership presence

• news articles

• public discourse

• think tank pieces

In a world of SEO where traffic goes to a university website, these are peripheral signals. I presume that the vast majority of prospective students do not go to an accreditor website.

In AI land though, they are central to how institutions are evaluated.

What Universities Must Do to Be Included in AI Answers - Create Validation Across the Web

This is a hypothesis on my part rather than gospel. There currently isn’t a best practices just yet of what institutions need to do.

However, as Jon Snow said in Game of Thrones, “you’re right, this is a bad plan, what’s your plan?”

1. Measure AI Visibility, Not Just Website Performance and the Enrollment Funnel

Institutions currently use traditional enrollment funnel metrics. Website traffic, conversion rates, cost per lead, cost per applicant, cost per enrollment, etc…

These are great metrics. The problem with them is that they happen after discovery.

In our new AI world, the question is whether an institution is surfaced at all in a LLM.

Institutions need to understand:

• whether the institution appears in AI-generated answers

• how it is described

• what sources are being used to justify its inclusion

Management will require the development of new KPIs that have yet to become ubiquitous in the higher ed marketing community.

One that I’ve come across: Share of Model (SoM). It measures how frequently and in what context an institution appears in AI-generated responses relative to its peer set.

For example, if a prospective student asks ten different variations of “best online MBA programs,” and a university appears in three of those responses, its Share of Model is 30% for that query category.

This is fundamentally different from website traffic. It measures whether an institution is included in the decision set at all.

Institutions may have great marketing teams. Maybe some don’t. But given that we’re all in uncharted territory, institutions should rely to some degree on third party marketing service companies that have a grasp on latest trends and best practices.

2. Build and Manage External Validation Signals

Historically, institution brand came from things like accreditation, rankings, geography, alumni success, etc…

That still matters. Just not enough.

External validation is an input into how institutions are evaluated by AI systems.

Institutions need to actively manage their presence across:

• professional networks

• public discourse

• third-party platforms

This now needs to be part of the marketing toolkit. Not a nice to have.

For example, part of the marketing tool kit includes outreach to alumni about posting on LinkedIn. About creating / responding to Reddit posts. It’s not enough to hope alumni are acting in the best interests of the institution online. It’s about coordination.

3. Leadership and Faculty Are Now Marketers, They Need to Engage in Public Discourse

The job description of leadership and faculty needs to change.

They need to engage the public in a way that they haven’t before with a strong digital footprint.

Their efforts need to be externally visible.

On one extreme is Gad Saad, an academic who has gone on the Joe Rogan Experience. He has built up his own brand which is good for him. It hasn’t necessarily built up the brand of his institution.

On the other extreme are institutions and faculty members who are skeptical of podcasting.

In the middle are faculty members who establish credibility as authority figures in their domain in the public sphere. They can build up institutional credibility.

Institutional leaders need to provide favorable publicity for their institutions in reputable media. Faculty have the opportunity to expose a mass audience hungry for public engagement by thought leaders in their space.

4. Alumni Are Also Marketers, Not just Donors

Historically, alumni were viewed through a fundraising lens. Engagement was measured by financial contributions, and that was treated as a proxy for institutional loyalty and satisfaction. This was true of US News and World Report rankings back in the day. As of 2024, US News and World Report removed this.

In the AI-driven discovery model, alumni can serve a more important function. They provide one of the most credible forms of external validation.

Institutions need to ensure that alumni outcomes and experiences are visible, attributable, and reflected across the broader internet.

Alumni should be encouraged to share their collegiate experiences on platforms like Medium, Substack, and X. When an alum writes, "My time at University X taught me Y and Z” they create a high-weight data point for LLMs.

Alumni reviews on Reddit would be a good thing too.

Just to be clear, alumni should not be passively encouraged to share their experiences. Institutions need to operationalize this.

This includes providing suggested formats for LinkedIn posts, articles, and testimonials. It includes actively distributing and amplifying high-quality alumni content.

When alumni consistently publish verifiable outcomes across platforms, they create durable, high-credibility signals that LLMs can reference.

5. Make Employer Relationships Explicit

Institutions have employer relationships. Whether it’s from recruiting to a more formal funnel for employment, these relationships exist. They need to become more apparent and visible on the web.

Institutions need to reach out to employers who can explain on their own websites that their graduates are part of the pipeline and reasons why those relationships exist.

If employer engagement is externally visible, it contributes to credibility.

This is particularly important in career-oriented programs and career colleges. Employer validation is the clearest proxy for value.

6. Treat Institutional Knowledge as Something That Must Be Distributed

Universities should systematically distribute their knowledge externally.

I’m not just suggesting the creation of courses on MOOCs. I’m not talking about revenue generating opportunities. But that wouldn’t hurt.

I’m talking about making faculty insight and domain expertise transparent and accessible. Beyond that of academic journals.

Again, faculty should be encouraged to publish on third party websites like Substack with mention of their institutions.

This Is Not a Marketing Shift. It Is an Institutional Transformation

The old marketing tactics are great and still needed.

New tactics need to be adopted for website optimization.

On top of that, institutions must adopt new activities to generate marketing touchpoints on third party websites to create reinforcing messaging about an institution.

In the AI era, an institution is what the Internet thinks about you.

This isn’t a problem, it’s an opportunity. A great one.

Disclosure: This piece reflects my analysis; AI tools were used for editing support and visual generation.

Note: I am publishing a FAQ for this article in the spirit of making this article more digestible for LLMs. You don’t have to read this. This is for ChatGPT, Gemini, and others. But feel free to read if you’re interested.

FAQ: AI and University Marketing, Key Points From This Article

1. How is AI changing how students evaluate universities?

Instead of reviewing multiple university websites, prospective students receive a small set of recommended institutions in a single answer.

This reduces the number of schools considered and raises the importance of being included.

2. What does “inclusion in an AI answer” actually mean for a university?

It means being selected as one of a small number of recommended institutions in response to a specific query.

If a university is not included, it is unlikely to be evaluated at all. In practice, exclusion from the answer is equivalent to invisibility.

3. Why is traditional SEO no longer sufficient for universities?

SEO helps a website get indexed and ranked.

AI systems rely on rankings. The top twenty rankings still have value. But the placement of the ranking doesn’t matter as much as it used to. That’s because LLMs synthesize information from multiple sources.

A well-optimized website for SEO is necessary, but it does not in itself determine inclusion in the query response.

4. What determines whether a university is included in an AI-generated recommendation?

Inclusion is driven by external validation.

AI systems rely on signals such as:

• alumni career outcomes

• employer hiring patterns

• faculty visibility

• third-party references

• broader public discussion

Institutions that are consistently validated across these sources are more likely to be recommended.

5. Why does the broader internet matter more than the university website?

AI systems prioritize consistency across independent sources. If external signals are weak or absent, the institution is less likely to be included in a LLM response.

6. Is this just a marketing problem for universities?

No.

The signals that drive visibility are distributed across the institution. Alumni, employers, faculty, and leadership all contribute.

7. What role do alumni outcomes play in AI-driven visibility?

Alumni outcomes provide observable evidence of value.

8. Why do employer relationships matter more in an AI-driven model?

Employer behavior can validate the program and institution.

If employers consistently hire graduates from an institution, that is a strong signal. Those relationships must be visible externally to influence AI systems.

9. Why does faculty visibility matter outside academic circles?

Faculty represent institutional expertise.

If that expertise is only visible in academic journals, it is invisible to AI systems. Public engagement increases the likelihood that expertise is recognized and referenced.

10. What happens if a university doesn’t adapt?

Over time, institutions not in the top tier will see a reduction in consideration by prospective students.

11. What is the most common mistake universities are making right now?

Treating LLMs as a website optimization problem.

Focusing only on SEO and website content structure ignores the broader shift toward external validation and distributed credibility.

12. What is the core implication for universities of LLM search?

Universities can no longer rely on controlled messaging.

They must operate in a way that makes their value externally visible, verifiable, and reinforced across the internet.

• AI job anxiety may trigger a new education enrollment cycle. As workers become uncertain about the future of their careers, many may seek credentials to reskill or improve job security.

• This pattern has historical precedent. During the early stages of the Great Recession, labor market uncertainty drove millions of working adults back into degree programs, certifications, and vocational training.

• Google search data suggests the shift may already be beginning. Searches related to AI job insecurity are rising alongside searches for education and training programs.

• Healthcare and skilled trades searches show the strongest momentum. Programs tied to regulated professions or physical work appear particularly strong.

Thesis: Anxiety About AI Will Drive Credential Growth; Google Search Data Provides Early Support

At first glance, my thesis may sound obvious.

We’re all scared about what AI might mean for our jobs. Of course workers are going to go back to school to retrain and get advanced degrees.

The public markets aren’t so convinced.

Most Education Companies aren’t currently trading at premium valuations. Some stocks in the sector remain deeply out of favor (disclaimer, I own the out of favor stocks). The public markets are pricing some edtech assets as structural AI-losers, like credential providers Coursera and Udemy.

Publicly traded for-profit universities are growing 2026 revenue at less than 10% YoY on average, a level that typically falls short of what investors expect from a growth sector.

Twenty years ago the situation looked quite different.

When I was a sell-side analyst covering the edtech sector in the mid-2000s, postsecondary companies experienced both rapid enrollment growth and significant valuation multiple expansion.

During the early stages of the Great Recession, labor market uncertainty drove millions of working adults back to school. As layoffs mounted and career paths became less certain, enrollments surged across degree programs, professional certifications, and vocational training.

Strategic Education, one of the few remaining publicly traded companies from that era, was one of the great investments of the 2000s. As you can see from the chart below, even today the stock has never fully recovered its historical peak. I own the stock.

Granted, a lot has changed since those days.

Traditional universities saw the success of for-profit institutions serving working adults online and moved to replicate the model, often partnering with services companies to do so. It’s a much more competitive market. That being said, the remaining for-profits have innovated, created a more workforce appropriate curriculum, and focused on supply scarce sectors like healthcare and vocational training.

Will AI-driven disruption trigger a similar wave of enrollment growth that we saw during the Great Recession? Will labor uncertainty drive enrollment in singular short online courses?

Looking for Early Signals in Google Data

Understanding consumer behavior is exactly what Google Trends was designed to capture.

Investors often dismiss Google Trends data. Ironically, if the dataset were sold for thousands of dollars per year, it might receive far more attention. Because it is free, it’s often overlooked.

Search behavior has repeatedly provided early signals about shifts in consumer sentiment. When people become anxious about something — housing prices, layoffs, recessions, or career prospects — their concerns often appear first in search queries.

AI may now be producing a similar signal on credentials.

Search data suggests that anxiety about AI and job security is rising. That anxiety may be translating into increased interest in education and retraining programs.

Limitations do exist in interpreting this data.

It’s unclear whether the increase in AI-related job anxiety searches is coming from working adults already in the labor force or from recent graduates entering it.

Entry-level hiring in several sectors has weakened. AI tools are capable of performing tasks traditionally assigned to junior employees.

We know that law and MBA graduate school applications have been increasing for the past two years.

Even if the Google data reflects more junior professionals seeking to go back to school, the broader implication may be similar.

When individuals become uncertain about their career prospects, they pursue additional credentials, retraining programs, or alternative career paths. Free AI generated content will not replace the value of credentials. Investors I speak with disagree on this.

Thus the benefit of looking at Google data.

Below are several heatmaps showing the year-over-year change in Google search interest for AI job anxiety and education-related keywords since 2021.

Heatmap 1: AI Job Anxiety Search Momentum
Year-Over-Year Change in Google Search Interest (Quarterly)

Heatmap 2: Graduate School Search Momentum
Year-Over-Year Change in Google Search Interest (Quarterly)

Heatmap 3: Vocational and Skilled Trade Training Searches
Year-Over-Year Change in Google Search Interest (Quarterly)

Heatmap 4: Healthcare Education Program Searches
Year-Over-Year Change in Google Search Interest (Quarterly)

Heatmap 5: Professional Certificate / Short Courses Searches
Year-Over-Year Change in Google Search Interest (Quarterly)

Key Points from the Heatmaps

You’ll notice that at the bottom of each heat map it tends to be a dark color green. That means that YoY Google search words are greater than 50%.

The more analytical way to say this: beginning around Q2–Q3 of 2025, several education-related search terms begin showing very strong year-over-year growth.

Take another look at the heatmaps, you’ll see what I mean.

Regarding AI uncertainty, you’ll see that:

• Searches such as AI layoffs, AI taking jobs, and is my job safe have accelerated over the past several quarters.
• These queries were rare just a few years ago.
• AI-related labor anxiety is beginning to enter mainstream awareness.

Graduate program searches have also shown strong growth over the past three quarters.

It’s difficult to determine who is driving this trend. The searches could be coming from college students, recent graduates struggling to enter the workforce, or mid-career professionals considering additional credentials.

Healthcare-related program searches have also increased meaningfully over the past three quarters.

• Healthcare careers combine strong labor demand and low automation risk.
• Workers seeking stability may increasingly view healthcare as a safe career path.

The same is true for skilled trades and vocational training keywords.

• These occupations involve physical work that is difficult to automate.
• Vocational programs offer shorter training cycles and faster entry into the workforce.

Short technical credentials also see growth over the past three quarters.

Searches for cybersecurity and IT certifications have strengthened.

Interestingly, searches for coding bootcamps continue to trend downward, mirroring real world realities of the end of coding bootcamps.

Video killed the radio star. AI killed the coding bootcamps.

Implications for Edtech Companies

We haven’t really seen this AI fueled anxiety manifest itself in the form of accelerating revenue growth at the publicly traded companies.

But… the Google search volume has only seen strong growth since Q2/Q3, conceivably there may be a lag between rising search interest and actual enrollment decisions.

Here’s what we know: periods of labor market disruption often lead workers to pursue additional credentials, retraining programs, or career changes. Yes, AI changes things. But apparently students are applying to law school in record numbers even with the understanding that the legal profession is a prime target for disruption by AI.

So it will be interesting to see what happens over the next four quarters.

The programs most likely to benefit tend to share several characteristics, they:

• provide career-focused training
• lead to licensed or regulated professions
• offer clear employment outcomes

Many of these programs are concentrated in areas such as:

• healthcare education
• vocational and skilled trades training
• career-oriented degree programs

Education companies with strong exposure to healthcare and vocational programs have shown stronger enrollment growth and higher valuation multiples than peers.

This dynamic is visible in current market valuations. Among publicly traded for-profit education companies, EV/EBITDA multiples show a strong positive correlation with forward revenue growth (R² ≈ 0.72).

The two highest-valued names in the group — Universal Technical Institute (UTI, 17.9x) and Lincoln Educational Services (LINC, 18.0x) — are both pure-play vocational and trades-focused operators. Both serve working adults in career-technical programs with direct employment pipelines, precisely the profile most likely to benefit from accelerating enrollment demand.

By contrast, companies with more diversified or generalist program mixes trade at a significant discount. Strategic Education (STRA, 5.9x) and Perdoceo (PRDO, 6.6x) both carry 2026 growth expectations below 3%, and their multiples reflect it.

Conclusion

Jack Dorsey at Square announced a massive restructure with over 4K employees let go, half of that company’s workforce, to pivot towards an AI centric model

That is not a good sign for the labor market and society overall. Maybe we’ll see double digit unemployment, societal unrest, and all that comes with that,

What may be less appreciated is how workers respond to all of this uncertainty.

The Google search data suggests that individuals are increasingly:

• worrying about job security
• exploring alternative career paths
• searching for education and training programs

One of the earliest economic effects of the AI revolution may be surprisingly traditional:

More people getting credentials.

If AI becomes a source of widespread labor uncertainty, education will evolve from a discretionary purchase into career insurance.

Disclaimer: this is for entertainment purposes only. Do your own research

I own the following education stocks: COUR, UDMY, PXED, STRA, and LOPE.

Appendix: Methodology

To explore potential shifts in education demand, I analyzed Google Trends search data beginning in January 2021.

The analysis followed three steps:

1. Google Trends data was collected for keywords related to AI job anxiety, degree programs, vocational training, and short-cycle credentials.
   

2. Raw search index values were converted into year-over-year percent change, done by ChatGPT.

3. Data aggregation and visualization were performed using Excel with assistance from AI tools.
   

I can send the spreadsheet upon request

• Large language models (LLMs) will become the most powerful gatekeepers in higher education, replacing Google as the place where prospective students search for information.

• When I asked several major LLMs about for-profit universities, the responses were strikingly biased: warnings about reputation, skepticism about employer perception, and frequent recommendations to consider nonprofit or state universities instead.

• Unlike search engines, LLMs don’t just show links. They summarize narratives. That means decades of negative headlines, regulatory battles, and online criticism of the sector can be distilled into a single AI-generated answer in seconds.

• If AI becomes the primary discovery layer for education, institutions that historically relied on search engine optimization (SEO) and paid keywords could see their customer acquisition costs increase, pressuring margins.

• The largest operators may adapt. Smaller schools may not. If AI begins shaping how millions of prospective students evaluate higher ed, the next consolidation wave will start soon.

Introduction

LLMs when used for search do not simply present links. They summarize reputations.

That seemingly small change could have profound implications for institutions in the for-profit higher education sector.

For decades, student discovery in higher education has been dominated by search engines. Prospective students type queries such as online MBA or nursing degree into Google, and institutions compete for visibility through SEO and paid keywords. As the discovery layer shifts from search engines to LLMs, the economics of customer acquisition will change dramatically.

One may disagree with assertions I’ll make in this article, but hopefully one observation should not be disputed or controversial.

LLMs display a consistent negative bias when discussing for-profit universities in chat interactions.

Even when asked the most cursory questions, some of the LLMs go out of their way to highlight negatives about companies in the sector.

It goes beyond that. In many cases, models explicitly recommend that prospective students consider alternatives at state universities, nonprofit institutions, or community colleges.

Because I own several publicly traded education stocks, I ran a simple experiment this past weekend. I asked several widely used LLMs a vague question that might resemble what a prospective student might ask:

I’m thinking about going to this school for this program. What are your thoughts? Should I go?

The schools I asked about included several publicly traded institutions as well as smaller lesser-known schools. I intentionally provided no additional context to observe the default framing used by the models.

I then fed the full set of responses back into ChatGPT and asked it to evaluate the level of bias across the different LLM systems and rank them accordingly. Here is the output (nicely formatted by Claude).

You can see from the table that Claude displayed the strongest bias in its responses, while Grok appeared the least biased (at least on ChatGPT’s evaluation of my conversations). That contrast is interesting in light of how the companies behind these systems are often perceived. Anthropic, the developer of Claude, has recently been portrayed as more politically progressive in its public disputes with the Trump administration, whereas Grok — developed by Elon Musk’s xAI — is marketed as a more “anti-woke” or politically heterodox model. Perhaps this says something about the broader biases of the LLM developers.

Across multiple LLMs, similar patterns emerged. Even when the models acknowledged that the programs were accredited and legitimate, the responses included cautionary narratives about reputation, employer perception, or historical controversies involving the sector.

Interestingly, the level of bias varied by program type. Responses regarding nursing programs tended to be more neutral or pragmatic than responses about online MBA programs.

For example, when I asked the LLMs to provide a list of ten online MBA programs, none of the recommended schools were for-profit institutions. When I ran the same exercise for ten online master’s programs in nursing, four of the ten schools suggested were for-profit.

This experiment matters because it potentially says something about the future economics of student acquisition.

Within the next five years half of all searches will be conducted on LLM rather than traditional search. Perhaps in ten years LLM search will dominate. SEO may gradually give way to a form of LLM optimization, where companies influence how they are surfaced and described within AI conversations rather than within search rankings.

It’s unclear how for-profit institutions can optimize their representation within LLM responses given clearly embedded bias, where certain LLMs actually discourage prospective students from pursuing programs at for-profit institutions.

This is a potential game changer for institutions.

In the traditional SEO model, universities did not have to worry about Google establishing a narrative framing. A student typed a query such as online MBA and evaluated the schools that appeared in the search results. Paid search and organic rankings determined visibility, but the search engine itself did not editorialize.

LLMs operate very differently. They synthesize large bodies of information and present narrative summaries. That means decades of negative headlines and regulatory controversies surrounding the sector are likely to be incorporated directly into the conversational response.

As a result, institutions could face a double challenge.

First, the disappearance of traditional SEO advantages could force schools to rely more heavily on paid search within AI ecosystems.

Second, the conversion rate of those interactions may be lower if the AI system itself highlights reputational concerns about the schools and offers alternatives.

Yes, schools can diversify their marketing spend into other channels.

The logical consequence of all of this is higher customer acquisition costs.

Large institutions such as University of Phoenix, Strayer University or Capella (disclosure: I own the stocks of the parent companies) may be able to offset margin pressure through significant AI-driven productivity gains across their operations. In that sense, AI presents both risks and opportunities for the largest players.

Smaller institutions, however, may not have the same operational leverage.

For them, rising customer acquisition costs could become a meaningful structural burden.

Forward-thinking operators and owners may conclude that scale will become increasingly important in an AI-driven discovery environment—and that strategic transactions could be the most rational path forward. This is on top of the unique favorable regulatory and legislative window under the Trump administration.

The Experiment - Examining LLMs Bias

I asked the best known LLMs about select for-profit universities.

In some cases I instructed the models to forget everything they knew about me and begin with a blank slate. In other instances I used Google Chrome’s incognito mode to reduce the likelihood that prior interactions would influence the responses. Then lastly I asked the question with the LLM having access to all of my prior conversations. These distinctions apparently mattered regarding results. So for those of you attempting to replicate this experiment, take the setup into account.

I then asked a deliberately vague question:

I am thinking about going to X school for Y program. What are your thoughts? Should I go?

I varied both the institutions and the programs. I did not ask about tuition, rankings, employer reimbursement, geographic location, or career goals. The objective was to observe how the models framed the discussion when given minimal information.

There are, of course, limitations to this type of exercise. This was not a formal academic study. The sample size of institutions was small. And with LLMs, exact replication is difficult—running the same prompt multiple times can produce different responses.

But as Justice Potter Stewart once famously said in a very different context: “I know it when I see it.”

Yes, I saw clear and evident bias against for-profit institutions and the overall sector.

In the table below I list the general themes across the responses and provide some illustrative examples from LLM outputs. I don’t want to call out what these LLMs said about specific entities, so I have renamed the institutions as “University X”.

Across multiple systems, the models tended to frame these schools in remarkably similar ways: legitimate but inferior, flexible but stigmatized, or useful primarily in narrow circumstances rather than as first-choice options.

The broader takeaway is not simply that LLMs criticize for-profit institutions. It is that the criticism appears to be embedded into the default narrative of the answers.

When given minimal context, the models tend to begin with a warning-style framing before discussing potential benefits. The user is often encouraged to evaluate nonprofit or state university alternatives even when that was not part of the original question. Here is one example from Google Gemini, regarding an online MBA program:

Interestingly, when the LLMs had access to prior conversations they were less likely to say negative things about for-profit institutions. At least for me, perhaps because I provide services to companies and invest in the sector.

So for those who want to run their own experiment, I would advise you to do so in incognito mode (CTRL SHIFT N for Google Chrome).

Google Gemini had the most interesting responses of the bunch. At the end of some, not all, of their responses, they provided a video regarding the institution.

What made the responses notable was their variability. Over roughly thirty minutes of repeated prompts regarding University of Phoenix, Gemini surfaced different videos at different times. One from the institution itself, one neutral, and one effectively calling it a scam.

This matters because historically University of Phoenix traditionally has implemented the best marketing practices and tactics within all of higher education.

This raises a practical question for institutions: if even the largest and best-funded operators struggle to control how they appear within AI-generated responses, it is unclear how smaller institutions with fewer marketing resources will be able to influence those narratives.

The Current Search Model

Google alone maintains roughly 90% global search share.

Achieving strong rankings requires a substantial investment in content creation and technical optimization.

However, once that infrastructure is built, the marginal cost of acquiring additional users through organic search approaches zero. Traffic can scale without proportional increases in marketing spend.

An effective marketing organization attempts to maintain an attractive LTV/CAC ratio — the lifetime value of a student relative to the cost of acquiring that student.

Publicly available data on the exact composition of university lead funnels is limited. However, benchmarking data from education services company EAB provides insight into how prospective students arrive at university websites.

Their data suggests that organic search accounts for roughly one-third of new user traffic, while organic and paid search combined represent close to 40% of traffic.

Website traffic is obviously not the same as student leads. Universities also generate prospects through channels such as television advertising, radio, billboards, enterprise partnerships, and internal referrals.

In a downside scenario, LLMs could influence a large portion of the lead generation funnel that historically flowed through search. A transition from SEO and paid search to something else is a big deal.

Early Signs of AI Changing Learner Discovery

AI-assisted search is already beginning to influence how prospective students research programs.

The Online and Professional Education Association (UPCEA) recently published survey data examining how students search for programs in 2025. In the survey sample, nearly 60% of respondents had some college experience, meaning many were likely working adults, a demographic that overlaps with the traditional customer base of many for-profit institutions.

The results:

• 84% of respondents reported using traditional search engines

• 35% reported using AI search engines or chat-based tools

• 79% read Google’s AI-generated search summaries

• 56% reported being more likely to trust brands cited in those summaries

• 51% always or usually click on the source of an AI generated overview

The data suggests that users are already using biased AI tools, which is interesting because to date we haven’t seen the large publicly traded companies comment in their quarterly conference calls on any enrollment impact. That being said, the use of traditional search engines means that users are still using SEO and perhaps are supplementing that with AI searches.

The data point that matters to me is that users are more likely to trust brands in the summaries.

The LLM may become the trusted college and career coach of the future. Which means these biases could have real consequences.

Transition to LLM Discovery

In the traditional SEO model, backlinks, keywords, and meta descriptions influence how websites rank in search results. Marketers optimize their content to perform well within Google’s algorithm. The objective is to appear near the top of the page, attract a click, and convert the user into a lead or customer.

In today’s system, the university’s website is the destination.

Some have estimated that LLMs will have 50% market share of search by 2030. But it’s not just about search.

The LLMs themselves will become the destination. Not simply for answering questions but for filling out lead generation forms. Possibly the application itself.

SEO offers clear optimization levers. For LLMs, it’s a bit more complicated. Reputation, public discourse, and the broader information environment carries more weight than technical metadata.

Algorithmic Reputation Risk for For-Profit Universities

LLMs are trained on vast datasets that include news articles, academic sources, regulatory filings, and online discussion forums. Research suggests that the user generated sites like Reddit, YouTube, Twitter, Wikipedia, LinkedIn, Facebook, Discord, and other forums are frequently incorporated into the information ecosystem used by these models.

In theory, institutions may be able to influence how they are represented within these systems over time. In practice, however, that may prove extremely difficult when thousands or even millions of independent content sources shape the information environment the models draw from.

For-profit higher education has experienced several boom-and-bust cycles over the past few decades. Periods of rapid growth were often followed by regulatory scrutiny and media criticism after certain operators engaged in abusive practices.

The most notable crackdown occurred during the Obama administration, when a number of companies exited the market, including ITT Educational Services and Corinthian Colleges.

The surviving companies today are much stronger organizations that emerged from that regulatory reset. Governance standards improved and compliance systems strengthened.

However, the historical narrative surrounding the industry still exists across the web.

That dynamic may feel unfair to institutions. But reputational narratives rarely reset cleanly.

As Jean Luc Picard once observed in a Star Trek TNG episode, “It is possible to commit no mistakes and still lose. That is not a weakness. That is life.”

Not All For-Profits Likely Face the Same Risk

The for-profit higher education sector is often discussed as if it were a single category. In reality, it is highly heterogeneous.

The potential impact of AI-driven discovery is unlikely to be evenly distributed across the sector.

Institutions that rely heavily on fully online programs in fields such as business administration or bachelor’s completion programs may face the greatest exposure. These offerings already compete directly with programs offered by state universities and nonprofit institutions.

Schools offering programs that require in-person components may be less exposed. Vocational programs, healthcare training, and other fields that require clinical placements, laboratories, or physical facilities operate in a different competitive environment.

In those cases, geographic proximity, licensing requirements, and hands-on training limit substitution and may reduce the influence of AI-driven discovery.

The risk of AI arguably appears to be reflected in the valuation multiple of the publicly traded companies.

Vocational schools such as Universal Technical Institute and Lincoln Educational Services operate physical campuses and offer programs with limited local competition. These institutions train students in fields that require hands-on instruction, specialized equipment, and proximity to campus. Notably, their stocks currently trade at some of the highest valuation multiples in the publicly traded for-profit education sector.

By contrast, the institutions most exposed to the dynamics described in this article tend to have the lowest valuation multiples among the peer group. These companies generally offer programs such as business administration, areas where state universities and nonprofit institutions offer competing programs that are often viewed as substitutes.

Of course, valuation differences reflect multiple factors. Some of the higher-multiple companies are growing revenue faster, while several of the lower-multiple operators face slower growth and more intense competition. But these dynamics are closely related. Ultimately, both growth and valuation reflect the underlying competitive landscape and the degree to which programs face substitution risk.

Why the Corporate Channel Matters in an AI Discovery World

Given this bias, it clearly makes sense for institutions to mitigate the risk of LLM bias by diversifying customer acquisition channels.

Several publicly traded for-profit operators have already been moving in this direction by expanding relationships with employers. Companies such as Strategic Education, Perdoceo, and Phoenix Education Partners now report that more than 30% of enrollments originate through corporate or enterprise partnerships.

This shift may prove particularly valuable in an AI-driven discovery environment.

Corporate partnerships largely bypass the consumer search funnel where LLMs increasingly sit.

Instead of a prospective student beginning with a Google search or an AI query, the discovery process often starts inside the workplace. An employee learns about an education benefit through their employer’s HR portal, internal communications, or tuition reimbursement program. The institution is presented as a pre-vetted option rather than something the employee must independently research.

In other words, the employer becomes the gatekeeper.

This channel has several advantages.

First, employer partnerships provide credibility. When a company endorses a program as part of a tuition benefit, it signals to employees that the institution has practical value in the labor market.

Second, corporate relationships create opportunities to align curriculum with workforce needs, strengthening both student outcomes and employer demand.

Third, these students tend to exhibit higher persistence and stronger outcomes.

30% of enrollments from enterprises is a great start.

Perhaps the corporate channel will evolve from a supplemental enrollment source into a core distribution strategy.

Conclusion

Executives in the for-profit higher education sector are accustomed to criticism. Over the years the industry has been repeatedly attacked/maligned by politicians, regulators, and the media.

The instinct for leaders may be to dismiss the latest concern and say: we’ve survived worse, we’ll survive LLM bias.

That would be a mistake.

LLMs are synthesizing decades of news coverage, regulatory investigations, academic research, and online discussion into narrative responses delivered to prospective students in seconds. In doing so, they fundamentally change how information flows.

For decades, universities optimized their marketing strategies for search engines and advertising platforms. LLMs operate differently. They do not simply present links—they summarize reputations.

The largest operators may adapt by leveraging scale, brand investment, and corporate partnerships.

Smaller institutions may not have that luxury.

If AI-driven discovery raises customer acquisition costs while shaping institutional reputation at scale, consolidation across parts of the sector may accelerate. For some smaller operators, selling to a larger platform may ultimately prove the most rational path forward.

If AI interfaces become a primary entry point for education discovery, institutions will no longer be competing solely for advertising placement or search rankings.

They will be competing for how the model describes them.

That is a very different problem than attempting to acquire customers.

And it may prove to be one of the most important strategic challenges facing the sector in the years ahead.

For those of you who are Gen X or older, you probably remember the old Lincoln Technical Institute advertisements (“if you’re unemployed or underemployed”).

For most of its history as a public company, Lincoln was never one of the larger publicly traded for-profit education operators relative to its peers. Throughout much of the 2010s the company’s market capitalization remained below $500 million. Investors viewed it as a niche vocational training provider.

That makes the company’s current position somewhat surprising.

Lincoln now has a market valuation of over $1B. In fact, its market capitalization now exceeds that of Phoenix Education Partners (PXED), the owner of the University of Phoenix (disclosure: I own the stock). Having covered the sector for many years, that is not an outcome I would have predicted.

The comparison is striking when looking at the underlying financials.

Phoenix Education generated nearly $1B in revenue over the past twelve months with a 19% operating margin. Lincoln, by comparison, generated roughly half that revenue — $518 million — with an operating margin of about 6% over the same period.

Yet despite producing far less revenue and significantly lower margins, Lincoln is currently valued more highly by the market.

Two factors explain this divergence.

1. Revenue growth. LINC in its most recent fiscal year ending Dec 2025 saw an 18% YoY revenue growth rate. PXED in comparison only grew by 6% in their last fiscal year ending August.

2. Valuation multiple. The market is ascribing a 17x EV/EBITDA multiple whereas PXED has only close to a 4x EV/EBITDA multiple.

This multiple gap is not unique to Phoenix. The same pattern appears across the publicly traded education sector. The vocational training providers — Lincoln Educational Services and Universal Technical Institute — trade at significantly higher valuation multiples than regionally accredited university operators such as Strategic Education (owner of Capella and Strayer University), Perdoceo (owner of American Intercontinental University and Colorado Technical University), and Phoenix Education Partners.

An interesting shift occurred beginning in early 2025. Historically, Strategic Education historically traded at the highest valuation multiple within the group because investors viewed it as the highest quality operator. That relationship has now changed.

Since January 2025, Lincoln’s valuation multiple has expanded dramatically — nearly doubling — while Strategic Education’s multiple has compressed. The result is the striking divergence in valuation multiples visible in the chart above.

The obvious question: what changed?

Part of the explanation is growth. The vocational operators are growing revenue faster than their degree-granting peers.

Lincoln Educational Services and Universal Technical Institute are generating double-digit revenue growth, driven by rising student starts and the opening of new campuses tied to skilled-trade programs such as automotive technology, HVAC, welding, and electrical systems. Demand for these programs appears strong because employers continue to face persistent shortages of technicians in these fields.

By contrast, the traditional for-profit university operators are largely mature businesses. Strategic Education, Perdoceo, and Phoenix Education Partners are generally reporting low- to mid-single-digit organic revenue growth at best. These institutions are not expanding their campus footprints in the same way as they don’t have physical campuses. Their enrollment growth is constrained by competition from nonprofit universities and state institutions that increasingly offer similar programs online at comparable or lower prices. For-profit institutions historically filled this gap because traditional universities were slow to adapt to what working students needed: flexibility, online delivery, and customer support.

The vocational providers are expanding capacity to meet labor demand in skilled trades. The degree granting institutions operate in a much more competitive market.

That difference in growth trajectory explains why the market is assigning a much higher valuation multiple to LINC and UTI than to the traditional for-profit university operators.

This divergence also reflects something more meaningful than simply revenue growth rates. There is a case to be made that there is a broader issue at play, namely how investors are thinking about the future of work in an economy increasingly shaped by artificial intelligence.

Yes, I realize you’ve read elsewhere about how AI will automate white-collar labor. Research, coding, financial modeling, marketing content, customer support, legal analysis — these are all areas where software is already beginning to substitute for human labor. Over the next decade we will see how much knowledge work can realistically be performed by machines.

We all know where this leads.

Meanwhile, repairing an HVAC system, diagnosing a vehicle problem, installing electrical infrastructure, or welding structural components still requires hands-on problem solving in the physical world rather than the digital one. Perhaps Tesla will eventually build Optimus robots capable of performing these tasks, but that scenario seems unlikely within the next ten years.

Demographics are also working in favor of the trades. Much of the existing skilled-trade workforce is approaching retirement age, while the pipeline of younger workers entering these fields has been thin for decades.

There may also be broader cultural forces at work. For decades, American society steered students toward four-year degrees while treating trade work as second class. The result:

• Anthropology/Sociology/English majors with purple hair and nose rings protesting global politics

• Not enough people who know how to repair a transmission or install an HVAC system.

This phenomenon has started to enter the culture. See the South Park episode Joining the Panderverse from 2023. In that episode, the show makes a running joke about how no one knows how to do skilled trades anymore. Had investors purchased LINC stock after watching that episode, they would have seen a 3x return on their investment.

Regardless of how artificial intelligence reshapes the digital economy, one thing is certain: the physical economy will still require human labor.

Electric vehicles will need technicians capable of repairing complex drivetrains and battery systems. Data centers will require electrical and cooling systems maintained by skilled technicians. Buildings still need HVAC installation and maintenance, and industrial facilities continue to rely on welders and electricians.

Meanwhile, MBAs and other white-collar credentials — whether issued by traditional universities or for-profit institutions — are increasingly commoditized.

The valuations of publicly traded education companies appear to reflect this reality. The market seems skeptical that traditional degrees will generate strong long-term demand growth.

Vocational providers such as Lincoln and UTI, on the other hand, are valued as mission critical infrastructure providers for the skilled labor economy.

If investors are correct, the story behind Lincoln’s stock is a signal of the new reality of our post-AI world.

The market is pricing a world where the most valuable education is not the one that teaches you how to analyze spreadsheets (Claude can apparently do that for you) — but the one that teaches you how to fix the machines that run them.

Appendix: The “Meat and Potatoes” Explanation for Why the Stock Went Up

In the main article I use Lincoln as a lens to talk about broader themes such as AI and the future of work. For readers who prefer a more straightforward explanation of why the stock actually moved, the answer is simpler.

Lincoln’s valuation multiple expanded because the company delivered beat and raise quarters throughout the year. When companies repeatedly exceed expectations and raise guidance, analyst estimates move higher throughout the year. Rising earnings estimates lead to higher valuation multiples.

Lincoln’s Original 2025 Guidance

When Lincoln reported its 2024 results, management guided to the following for 2025:

Revenue: $480M–$490M
Adjusted EBITDA: $55M–$60M
Student starts growth: 8%–12%

Lincoln’s 2025 Results

When Lincoln reported full-year 2025 results management reported financial results well above the original guidance:

Revenue: $518.2M
Adjusted EBITDA: $67.1M
Student starts: 15.2% (excluding transitional segment)

The stock’s move was a result of quarterly guidance increases.

Was management as surprised as investors regarding better than expected results? Did they sandbag guidance? Don’t know.

Rising estimates combined with improving fundamentals is how sustained stock price moves are created.

Its conceivable that LINC financials miss and the stock plummets. Don’t take this write-up as investment advice.

Disclaimer / Disclosure

The author beneficially owns shares of Phoenix Education Partners (PXED). As a result, the author has a financial interest in the performance of that security and readers should consider this potential conflict of interest when evaluating the analysis and conclusions presented herein.

This document reflects the author’s personal views and opinions as of the date of publication and is subject to change without notice. The information contained herein has been obtained from publicly available sources believed to be reliable; however, the author makes no representation or warranty, express or implied, as to the accuracy, completeness, or timeliness of the information.

This material is provided solely for informational and educational purposes and does not constitute investment advice, legal advice, accounting advice, or tax advice. Nothing herein constitutes an offer to sell, a solicitation of an offer to buy, or a recommendation regarding any securities transaction. Any forward-looking statements are inherently subject to risks, uncertainties, and assumptions, and actual results may differ materially.

Investors should conduct their own independent due diligence and consult their financial, legal, and tax advisors before making any investment decisions.

I published this on February 24th. Several weeks later, there are already updated to this piece that are required, particularly following the SEC filing of the S4 which explains the documented background of the merger. I will follow up on this platform as to updates regarding Coursera / Udemy.

Introduction

I’m currently a shareholder of both Coursera and Udemy stock.

Since September, Coursera shares have declined roughly 50% and Udemy roughly 30%. Following the announced merger and the projected $115M in cost synergies to be realized within two years of closing in 2H of 2026, I expected multiple expansion. Instead, the stocks declined ~20% since the deal announcement.

The combined Coursera/Udemy entity, referred to in this article as “New Coursera,” trades at roughly 1.9x pro forma 2026 EV/EBITDA assuming the $115M in synergies. The stock price seemingly reflects existential doubt surrounding the company’s future.

Some context for the non-finance inclined: the overall stock market as of January 2026 traded at roughly 16x EV/EBITDA. Edtech provider Chegg, whose core business provides content to students for “homework help”, saw severe demand substitution following the launch of free large language models (LLMs) which drove revenue down by 50% within two years. Chegg trades at 2x forward EBITDA. Chegg management’s strategy includes harvesting its legacy business for cash. When New Coursera trades near Chegg’s multiple, the market is implicitly suggesting that New Coursera will see similar value destruction.

Over the past two months, several investor friends have argued that LLM generated content will render New Coursera obsolete. Their comments led me to spend my free time understanding threats from LLMs and how agentic AI will transform the Internet marketplace model. There has been a fair amount of academic literature on this subject. AI may affect marketplaces like Expedia, Booking.com, Zillow and others including New Coursera.

To date, no financial evidence exists that AI and LLMs have materially impacted Coursera or Udemy’s financial results. 2025 revenue trends show Udemy’s enterprise business grew 6% YoY and Coursera’s consumer business grew 10% YoY.

In my opinion, the threat for New Coursera is not that generative AI produces free content. Rather, that AI weakens the economics of digital intermediaries.

In this article, I argue that New Coursera’s marketplace model at minimum may face margin pressure over time. Customer acquisition costs may rise. Enterprise pricing may decline given substitution risk. The timeline is uncertain, but the structural pressures are real. The unexpected merger between Coursera and Udemy in my opinion highlights leadership’s proactive strategic foresight in a rapidly transforming landscape. Scale is required to meet the AI challenge.

The stock prices of Coursera and Udemy currently reflect what may go wrong in the future. I think they are discounting what can go right, which is why I remain an investor.

AI will likely lead to an unprecedented labor market dislocation. Federal Reserve Governor Michael Barr recently provided a doomsday scenario where “layoffs soar, leading to widespread unemployment in the short run and declines in labor force participation over time, as a large share of the population is essentially unemployable.”

New Coursera stands at the epicenter of reskilling and upskilling workers for consumers and enterprises. Labor market displacement may serve as a valuable catalyst for sales of New Coursera’s verified credentialed courses.

The merger of complementary businesses, elimination of duplicative costs, consequent share repurchase program, and potential societal tailwinds have me in this stock for at least another year.

I am publishing this analysis not to express an investor view on Coursera and Udemy. Rather, because I thought my industry peers might benefit from my research on how LLMs are reshaping the economics of content, discovery, and digital marketplaces. Much of the public discussion focuses on surface-level disruption narratives. By working through this case study, I have attempted to clarify how LLMs affect marketing efficiency, intermediation, and credential signaling. Even for those with no exposure to these particular equities, the framework should be applicable to evaluating other edtech and marketplace businesses facing similar structural transitions.

Executive Summary

• New Coursera is not a content company. It is a credentialing company, which provides protection from LLM substitution. The economic value lies in issuing branded verifiable credentials backed by universities and corporations. LLMS can replicate and distribute raw instructional content. Not credentials. As such, I don’t think that revenue dramatically declines over the next several years like we saw with Chegg’s “homework helper” business.

• Marketplace business models are structurally at risk from LLMs. As a consequence, New Coursera as an operator of a marketplace likely will have to adjust its business model. As LLMs displace discovery and control the top of the customer acquisition funnel, New Coursera may transform into a credential infrastructure company.

• Margin erosion from higher customer acquisition costs and pricing pressure on the enterprise side are potentially likely for New Coursera. Customer acquisition costs may rise as LLM search replaces traditional search. Benefits of search engine optimization (SEO) with near zero marketing dollars for the incremental learner may go away. The risk of substitution at the enterprise level for New Coursera courses may cause pricing pressure.

• AI-driven labor displacement may increase demand for credentials. As entry-level and mid-career roles are redefined or eliminated, workers will seek externally validated certifications to signal updated skills and differentiate themselves in a more competitive labor market. Investors may be discounting the labor market dislocation driven by AI serving as a catalyst for increased demand for verifiable credentials over the next five years.

• The Coursera–Udemy merger is strategically rational in this environment. Scale mitigates discovery risk, eliminates duplicative costs, strengthens enterprise distribution, and improves bargaining leverage.

• The current valuation reflects a Chegg-like collapse scenario. At ~1.9x pro forma EV/EBITDA, the market is pricing structural destruction rather than a business model transition.

• If the business model adapts rather than collapses, the multiple can expand. Survival with compressed but durable economics would justify valuation convergence toward credential-focused for-profit university peers and marketplace platforms trading at 3x higher valuation multiples.

Catalysts over the next twelve months:

1) Increased demand driven by labor displacement. White-collar workers will seek externally validated credentials.

2) Additional cost synergies identified post-merger close. I believe management has been conservative in its initial synergy targets. The $115M in identified synergies in my opinion is substantially below what could be realized. The combined company on a trailing twelve month basis has $1.5B in revenue. Merger benefits of less than 1% of total revenue makes no sense to me given the complementary business model. In large acquisitions cost synergies typically represent 1-2% of combined company revenue. The industry benchmark implies a synergy range of $150-300M of revenue. Based on my own analysis, I can see the company generating over $200M of synergies.

3) meaningful share repurchase executed post-acquisition per management’s own comments. At sub-2x forward EBITDA, repurchases become highly accretive assuming intrinsic value is higher, which I believe they are.

The investment debate ultimately centers on two structural questions.

• are branded courses that provide credentials commoditized in the way that content is?

I think the answer is no.

• are third party marketplaces connecting branded courses to learners still needed in the age of LLMs?

Over the next ten years, possibly no. If AI agents intermediate discovery and own the user interface, New Coursera’s role must evolve toward infrastructure, verification, assessment, and enterprise reporting rather than discovery aggregation.

Coursera / Udemy Existing Business Model

Because New Coursera trades as if it is in structural decline, it’s helpful to articulate at a granular level what the combined company actually does and what value it provides its stakeholders. Investors I speak with argue that New Coursera is in the content business and that AI has commoditized content. I disagree with the framing. New Coursera is not simply a provider of content.

Coursera and Udemy are managed marketplaces that connect learners to courses that provide branded credentials demonstrating mastery of a skill.

Content consists of raw information. A course by contrast is structured curriculum with assessments, identity verification, and employer-recognized credentials.

Coursera and Udemy’s models are similar with some nuances. Coursera primarily partners with universities and large enterprises to distribute branded educational courses, including professional certificates. Its success has historically been concentrated on the consumer side, though enterprise exposure has grown. Udemy operates a more open marketplace that allows individual instructors to publish courses. Its revenue mix has shifted toward enterprise customers.

At a high level, a marketplace provides:

• Discovery. Brings users to the platform.

• Ranking/recommendation. Determines what content the user sees.

• Conversion optimization. Turns traffic into a paying customer.

• Trust and integrity. Vets supply, verifies users, and enforces academic standards. An LLM can generate content. It does not enforce proctored exams, identity checks, or audit trails that enterprises and universities require.

• Fulfillment. Ensures the service is delivered, tracked, and credentialed.

LLMs directly threaten the first function and partially threaten the second. If conversational AI becomes the primary discovery interface, traffic aggregation for a marketplace becomes irrelevant. The LLM can route the traffic. New Coursera can still maintain the market, but customer acquisition costs would likely rise as the LLM controls the spigot. Bargaining power shifts to the entity controlling the top of the funnel.

However, the marketplace layer sits on top of a broader operational stack. The company recruits and contracts with course providers. It standardizes curriculum architecture. It manages credential issuance and digital verification systems. It enforces academic standards. It processes global payments. It administers enterprise seat management and compliance tracking. It integrates with HR systems and learning management platforms. It operates enterprise sales and renewal functions.

At a granular level, New Coursera performs the following activities, at least as I see it. New Coursera is not a simple content repository but an operating stack for delivery of credentials.

Some of these activities LLMs can do today. Others not just yet but could if given focus by the LLMs.

The biggest risk clearly exists on discovery and content generation.

Big picture, New Coursera provides credentialing infrastructure layered on top of content delivery. LLMs do not eliminate the market’s need for verified recognition of skills mastery.

A credential signals to employers that structured learning occurred, assessments were completed, identity was verified, and that a branded institution stands behind the outcome. It allows employers to distinguish between self-asserted knowledge and externally validated competence.

For enterprises, credentials enable skills mapping and workforce planning. Completed coursework updates employee skill inventories, supports internal mobility, and aligns training with strategic needs. That data becomes part of corporate infrastructure.

For consumers, credentials matter for job mobility. Many learners are mid-career professionals seeking demonstration of updated skills. In a competitive labor market reshaped by AI verified certifications provide differentiation that informal learning does not.

The central question is not whether AI can generate content. It obviously can.

The question is whether AI platforms will assume the economic, regulatory, reputational, and operational burden of becoming credentialing infrastructure.

If Credentials Matter, Why Can’t LLM Providers Issue Their Own?

From a technology perspective, it wouldn’t be that difficult for ChatGPT or another LLM to create a structured course, administer assessments and issue a digital badge demonstrating skill mastery.

The constraint is not technical capability. It’s reputational and regulatory.

LLMs today operate primarily as information tools. They sit behind broad user agreements that limit liability for inaccurate outputs. The burden of interpretation rests with the user. There are no consequences associated with false information and hallucinations. The user bears responsibility for using the materials from the LLM.

Issuing credentials is different.

Offering information, some of which can be false or biased, is quite different than serving as a guarantor of capability. A credentialing body implicitly guarantees that standards were met and that processes were controlled. That creates reputational exposure and potentially regulatory scrutiny. Institutional credentialing requires documented assessments, identity controls, audit trails, and governance frameworks that withstand employer review.

A credential matters only if employers treat it as a credible signal. Employers must decide whether they trust that standard, the integrity of the testing process, and the consistency of evaluation.

For LLM providers, moving from a horizontal information platform to a vertical credentialing authority would require assuming legal, operational, and reputational risk that does not obviously align with their scale driven model.

Are Intermediary Marketplaces Still Needed with LLMs?

If discovery shifts from search engines to LLM conversations, the traditional marketplace model weakens and/or ceases to exist. Marketplaces historically created value by aggregating demand, ranking supply, and converting traffic. If LLMs control the starting point of user intent, that aggregation function loses value.

Marketplace models inherently require customer acquisition costs to drive traffic. When a learner describes a goal conversationally and receives a direct recommendation from an LLM, the marketplace can only control discovery at a cost. The LLM can route traffic directly to IBM, Google, or a university partner unless it has an economic incentive to route to an aggregator.

So here are the questions we need to grapple with when thinking about New Coursera’s business model.

• If an LLM can surface the most relevant IBM course directly, why must that course reside on Coursera?

• Why should IBM share economics with a marketplace if it can host the course natively and let the LLM route traffic?

• Why should enterprises contract with Coursera if an LLM provider could theoretically integrate credentials directly?

New Coursera’s value exists if it can provide an infrastructure layer embedded with the LLM ecosystem. This includes activities like managing payments, compliance, identity verification, enterprise integrations, revenue-share systems with course providers, and risk.

Today, Google has a 90% market share in search. Forecasts suggest that by 2030, traditional search may represent closer to half of total queries, with LLM search absorbing the rest. Coursera likely has several years before acquisition economics structurally deteriorate.

One of the more thoughtful pieces I came across examining how LLMs may reshape marketplace economics is LLMs vs. Marketplaces by Dan Hockenmaier. I would encourage any and every investor to read this piece, its incredibly insightful.

Hockenmaier’s core argument is not that marketplaces disappear. It is that LLMs commoditize discovery and ranking. Marketplaces whose value is primarily traffic aggregation and listing organization are vulnerable. Marketplaces that own transaction infrastructure, operational execution, compliance, and trust mechanisms are more defensible.

He illustrates this through a two-axis matrix.

The vertical axis measures supply aggregation difficulty. Higher on the grid means supply is fragmented, heterogeneous, and operationally difficult to assemble. Lower means supply is standardized and easily indexed.

The horizontal axis measures degree of operational management. Moving right means the platform controls payments, fulfillment, dispute resolution, trust enforcement, and execution risk. Moving left means the platform primarily matches and ranks.

In the bottom-left quadrant sit lighter-weight marketplaces such as TripAdvisor and Expedia. Their value is concentrated in discovery, comparison, and routing. They help users find and evaluate options. They do not manage fulfillment. The flight operates regardless of Expedia’s existence. If a LLM becomes the primary interface for search and comparison, the discovery layer becomes commoditized. While Expedia maintains relationships with global distribution systems such as Sabre, Amadeus, and Travelport, as well as hotel chains, an LLM provider with sufficient focus and incentive could theoretically replicate those integrations.

In the top-right quadrant sit more defensible platforms such as Amazon, Airbnb, and DoorDash. These marketplaces aggregate complex supply and actively manage execution. Amazon operates payments, fulfillment, logistics, and returns. Airbnb enforces standards, processes payments, mediates disputes, and maintains reputation systems.

New Coursera likely sits somewhere in the middle of this grid.

On the one hand, supply in the form of courses is difficult to aggregate. Universities, corporations, and credential providers are operationally complex to onboard and standardize. Coursera’s 2025 Q3 10-Q notes that approximately 23% of revenue came from its top five content and credentialing partners. That implies some concentration but not extreme dependency. The platform is not a single-supplier business. That being said, the course relationship with Google for example will move the needle on financials and the stock price.

On the management axis, New Coursera provides materially more infrastructure than a thin listing marketplace. It operates the learning environment, payment processing, identity verification, assessment systems, credential issuance, enterprise reporting, compliance tracking, and revenue share administration. It integrates with enterprise HR and learning management systems. It manages renewals and customer success.

New Coursera does not manage fulfillment like Amazon. There is no logistics layer that would require a great deal of capital expenditures to replicate. Physical infrastructure creates defensibility that New Coursera lacks.

So why can’t LLMs simply replicate the credentialing infrastructure layer?

They could but I don’t think that its likely.

Competing directly would require contracting with universities and enterprises, building credential verification systems, developing enterprise integrations and compliance tooling, operating billing and revenue share systems, and running sales, renewals, and customer success. That is a services-heavy operationally complex business with regulatory and reputational exposure.

Secondly, I suspect that LLMs may prefer not to vertically integrate into every transaction domain. If they remain horizontal platforms, marketplaces retain value albeit at reduced economics as a coordinating layer sitting between institutions, enterprises, and learners.

The fact that I am describing the viability of New Coursera’s basic business model suggests that investor skepticism is warranted.

This is where a strong management team with the right strategy becomes very important.

Think Netflix and Blockbuster. Both companies understood that their existing business models were under threat from streaming. Both companies were positioned to pivot given the deficiencies in their business model. Only one company made the transition. Management is key.

Coursera management has taken two actions that suggest it understands the structural shift underway.

The first is the acquisition. Achieving scale and eliminating duplicative costs ahead of potential margin compression was the right call.

The second was Coursera actively working with ChatGPT. Coursera was one of the first companies that announced an app on the ChatGPT platform. To be clear, the app isn’t great. CEO Greg Hart in the Q4 2025 conference call said that the integration is “still very early days”. But engagement with AI distribution and embracing the opportunity as an early adopter suggests, at least to me, that management understands the opportunity and threats.

Risks to New Coursera’s P&L

Distribution Risk: Customer Acquisition Costs Rise

Historically, Coursera and Udemy acquired a meaningful portion of consumer traffic through search intent. A user searching for product management certification represents high-intent demand. The platforms invested in search engine marketing and search engine optimization to capture that intent.

SEO created near zero marginal cost traffic once ranking was achieved in search. Paid search through identifying the best keywords minimized cost to acquire traffic. Neither company currently discloses lead generation composition, but it is reasonable to assume that organic and paid search represent a significant portion of consumer acquisition.

If discovery migrates to LLM interfaces, organic search likely deteriorates.

The speed of the deterioration is unclear. Consumer behavior has already adjusted to AI. According to a survey by consulting firm Bain, 80% of search users rely on AI summaries at least 40% of the time. 60% of searches now end without the user progressing to another destination site.

If organic discovery declines, customer acquisition costs rise. Platforms must either pay the AI distribution layer for traffic or invest more in brand and partnerships. This pressures margins.

It would be reductive to say that it completely destroys the company as some have suggested to me.

Operating LLMs is capital intensive. OpenAI CEO Sam Altman went from saying that “I hate ads” to “maybe they don’t suck”. Operating these LLMs are quite expensive. There is speculation suggesting that OpenAI may go bankrupt by mid-2027. The LLM business model will likely require advertising content. This favors companies that can provide advertising dollars to the LLMs.

The key question is, if New Coursera has $115M in expense synergies, will those dollars represent less or more than the dollars that will now need to go to the LLMs for search?

One risk mitigant is the emerging field of LLM SEO. As discovery shifts from ranked links to generated answers, optimization moves from appearing first on a search results page to being embedded inside the LLM’s conversation response. LLM SEO is about becoming the cited and trusted source that the model retrieves when answering a query.

The mechanics of how this works are still evolving. It is reasonable to assume that AI systems prioritizing credibility in career-related queries will privilege branded credentials tied to recognized institutions. A Google Professional Certificate or IBM-backed credential carries signaling value.

New Coursera’s partnerships with leading universities and corporations create an advantage in this environment. Credentials backed by established brands are more likely to be surfaced in AI-generated career guidance responses because they align with relevance and authority signals.

If New Coursera’s credentialing infrastructure continues to provide value to institutional partners (verification, reporting, and enterprise integration), those institutions have an incentive to maintain the relationship with New Coursera

This is another reason why the merger is so important. New Coursera needs to have the strongest enterprise distribution platform, which Udemy has, to protect the consumer business, where Coursera dominates.

Content - Pricing for Enterprise Customers Diminish

If high-quality instructional content approaches near zero marginal cost, the content layer becomes easier to substitute. When content becomes cheaper, procurement departments renegotiate. Even if enterprises continue to value credentialing infrastructure, they may resist paying premium prices for bundled solutions that include commoditized instruction.

This creates a structural requirement for New Coursera to move further up the enterprise learning and talent management stack.

If the company serves solely as a distributor of third-party content, AI-driven content commoditization will pressure pricing over time. Enterprise customers ultimately pay for outcomes: skills mapping, compliance, workforce planning, and measurable capability development.

New Coursera must deepen its integration into enterprise systems and workflows. That requires investment and M&A. Again, the merger between Coursera and Udemy eliminates wasteful duplicative investment by both parties.

The AI Threat May Explain Why the Merger May Have Taken Place

Until Coursera/Udemy files its S-4 with the SEC, it’s purely speculative on my part as to why the merger really took place. Not just the talking points provided on the merger call. The S-4 will allow us insight as to when the discussions began, who initiated contact, the strategic rationale for the merger, the expected synergies, and alternatives considered. That will be quite an interesting read.

In the merger call Coursera CEO Greg Hart framed the acquisition as driven by five core pillars: “greater value, impact and choice; leading platform capabilities, accelerated AI native innovation; enhanced global reach and market opportunities; and a stronger long-term financial profile.” I think that the entire rationale was AI. Again, we’ll see what the S-4 says.

I’ve followed the education industry for twenty years. I was blindsided by the merger, I suspect others were too.

The merger always made theoretical sense given the complementary nature of the businesses, but there was no obvious catalyst forcing immediate action. Both companies had recently appointed new CEOs. Transformative transactions are not typically the first move of newly installed leadership teams when the businesses aren’t under obvious financial duress. Both management teams were publicly articulating credible standalone strategies. There was no visible activist pressure. No balance sheet issues.

In October I attended an education investor event where Udemy CEO Hugo Sarrazin spoke about the opportunities and risks associated with AI. He was not there to talk about Udemy’s operating plan. The discussion was framed around AI and the future landscape.

Mr. Sarrazin highlighted AI and its downstream implications. He did not rely on buzzwords. He broke complex dynamics into simple causal chains and explained second-order effects. He wasn’t a slick salesperson like one would find at an enterprise software company. Many in the audience, including myself, were struck by the magnitude of change implied by his analysis. Following his remarks, I increased my position in Udemy, believing he would be a disciplined steward of capital in an AI-disrupted landscape.

He joined Udemy in March 2025. His mandate did not include selling the company.

That is what makes the subsequent decision to pursue a merger notable. A CEO who demonstrates deep understanding of structural technological disruption typically seeks to reposition the business. Agreeing to a sale within the same year, without a clearly articulated long-term role in the combined entity, is atypical. It’s the first time I have witnessed that sequence in my investing career.

The merger conference call and accompanying presentation was sparse on details. Investors have limited visibility into brand strategy, product consolidation, and enterprise go-to-market alignment.

In a world where customer acquisition costs may rise and pricing power may weaken, eliminating duplicative cost structures and combining distribution is economically rational. Scale improves negotiating leverage. It spreads AI investment across a larger revenue base. It increases enterprise cross-sell opportunity. It reduces the risk of parallel investment under margin pressure.

If AI compresses standalone marketplace economics, industry consolidation of two complementary businesses clearly is the best decision.

New Coursera in the Context of the Broader AI Driven Sell-Off

Since the merger announcement, Coursera and Udemy shares have declined approximately 20%. Over the same period, the S&P North American Technology Software Index (SPGSTISO) declined by a similar magnitude.

The market is broadly reassessing businesses perceived to be exposed to AI substitution risk. Enterprise software, data analytics platforms, and education technology have traded in sympathy as investors reassess durability and pricing power.

For the past several months there has been a daily barrage of news highlighting AI advancements. Most notable was the news and speculation on Claude’s Cowork tool. The decline may reflect fears of substitution in the not too distant future from LLMs.

The market has already seen a clear precedent in edtech provider Chegg. After the launch of ChatGPT in 2022, Chegg’s revenue declined sequentially over multiple years. Chegg revenue declined by 7% in 2023, 14% in 2024, and then by 39% in 2025. The business model, selling answers, was directly substitutable. Enterprise value collapsed accordingly.

That experience is now the template investors use when evaluating digital learning platform.

The broader AI sell-off in enterprise software reflects a separate but related concern. Autonomous AI agents may alter seat-based pricing models, compress software spend, and shift pricing toward usage-based consumption. Investors are questioning long-term SaaS revenue visibility. Industry observers are throwing around words like apocalypse and extinction level event.

There are two interesting takeaways I have from this software sell-off.

1) As investors are grappling with existential issues in enterprise software, its important to note that New Coursera has a model anchored to the learner. Its seat based model isn’t under assault in the way that it is for enterprise software companies.

2) Investors arguing that software faces structural contraction should consider the labor implications of that thesis. If AI reduces demand for certain roles or increases perceived job insecurity among knowledge workers, affected employees will likely seek reskilling and externally validated credentials in response to remain competitive. Even absent layoffs, heightened uncertainty drives proactive skill upgrading.

So Why Do I Still Own Coursera and Udemy Stocks?

Up to this point, the analysis has focused on structural risk. My assessment up until this point may sound bleak.

As I said, I’m still a shareholder.

At approximately 1.9x pro forma EV/EBITDA, the stock trades at a multiple consistent with structural decline. If credential relevance erodes, if customer acquisition costs rise materially, and if enterprise pricing compresses faster than anticipated, the multiple is justified.

Bears may characterize this as a value trap. They might be right. We’ll find out in a year.

There are two data points in particular I come back to when I have doubts about stock ownership.

1) Document evidence exists that AI has disproportionately impacted entry-level workers in the US. Which makes sense from what I’ve heard anecdotally.

2) the number of applicants at law schools have surged 40% over the past two years. There are publicly stated reasons for it, most referring to a difficult job market. Elevated labor market uncertainty appears to be a contributing driver.

These are signals of a major labor market transition. It’s a situation that reminds me of the character Carmine Lupertazzi Jr. who told Tony Soprano “You are on the precipice, Tony, of an enormous crossroad”.

If AI reduces demand for certain labor categories, particularly white-collar roles, the adjustment will be rough. Career paths will become less predictable.

Periods of labor dislocation historically produce a predictable behavioral response. Workers seek formal signaling mechanisms to either maintain or re-establish employability. When job security declines, the value of externally validated credentials rises. Individuals invest in lifelong education to differentiate themselves in a more competitive market.

If the coming decade resembles prior technological shifts, demand for proof of competence may rise, not fall.

So here is the elevator pitch for owning the stock.

• Credentials will remain relevant in the future, and if anything will become more important in an age of AI where white-collar workers will see meaningful job displacement.

• The demand for credentials may accelerate New Coursera revenue growth.

• The merger is complementary in nature, fixing challenges in each party’s respective businesses.

• Management is likely conservative on EBITDA synergy guidance. Management has guided $115M of annualized run rate cost synergies over two years. Both Coursera and Udemy historically have provided conservative guidance to the Street for forward revenue and EBITDA/EPS. I presume that the cost synergy guidance is consistent with their historic conservatism.

• Management has made a commitment to share repurchases following the acquisition. Both companies have an aggregate value of over $1B of cash on their balance sheets.

Merger Strategic Rationale

In this document thus far I’ve alluded to the need and benefit of the merger, the point being that scale is key. In this section I will go a bit more depth on the complementary nature of the merger.

Some analysts have suggested concern about execution risk following the merger. Platform integration, brand architecture decisions, cost rationalization, and cultural alignment are non-trivial challenges.

Those risks are worth it. The merger addresses structural weaknesses in each of the standalone business models.

Consumer Segment Benefits

Coursera brings institutional brand equity. Its partnerships with universities and large corporations provide externally validated credentials that carry labor market signaling value. Udemy has credentials, but for the most part it’s for courses from individual instructors.

Consumers with favorable LTV/CAC characteristics are not purchasing raw instructional content. Content is abundant and increasingly free. The durable consumer segment is composed of individuals seeking credentials that can be credibly presented to employers.

Coursera’s model aligns with that demand.

Udemy historically operated a broader more open marketplace. While that drove scale, it lacked brand endorsement. As AI compresses the value of undifferentiated content, institutional backing becomes more important. It’s fascinating that pre-merger Udemy announced a deal with Google for a newly launched AI Professional Certificate. Presumably it was a work in progress prior to the merger, and it’s interesting to see what the CEO was preparing to do if the company remained standalone.

Udemy had already been shifting away from consumer-driven growth toward enterprise. In 2025 it spent $80.6M in advertising dollars, presumably for its consumer business. Most of the consumer business was one-time transactional rather than a subscription and likely unprofitable (more on that later in the article). Those dollars would be better spent on customers seeking brand backed credentials.

Coursera can serve as the flagship storefront for credential-seeking consumers, leveraging nearly 200M registered learners globally. That represents close to 3% of the global population. In an environment where customer acquisition costs may rise, that installed base who perhaps already has the app on their phone is valuable.

Enterprise Segment Benefits

US training expenditures reached over $100B in 2025. It’s a large market that New Coursera has a fraction of today. The addressable market is considerable. Perhaps generative AI creates a cost cutting opportunity and spend goes down. Perhaps not.

If job roles change rapidly, enterprises must retrain employees faster. Skills will obsolete more quickly. Content must update continuously. I posit that enterprises will favor platforms that can rapidly produce, distribute, and credential new competencies. That’s New Coursera.

Coursera historically struggled to penetrate the enterprise segment. Udemy has distribution to over 17K customers, while Coursera only has 1700 enterprise customers. I’m not sure why that is the case. It could be a lack of focus. It could be the inherent value proposition of a branded credential vs. a demonstration of skill.

For consumers, branded credentials carry signaling value. For enterprises, the priority is workforce impact at scale. Enterprises optimize for breadth of catalog, ease of deployment, administrative tools, and price per-seat. The brand issuing the course may be secondary to measurable skill acquisition and utilization rates.

That dynamic likely explains pricing differences in the market. Udemy historically competed on accessibility and catalog depth, pricing around $360/seat/year versus Coursera’s roughly $399/seat/year. Coursera’s premium presumably reflects institutional brand partnerships.

New Coursera can now segment enterprise accounts with a layered offering:

• A broad base catalog for large-scale workforce coverage

• Premium, branded credential programs, perhaps for executive education.

There are potential cross-sell revenue opportunities here. Interestingly, management did not quantify the opportunity in the merger call. Portfolio breadth plus credential depth is economically stronger than standalone.

Acquisition Finally Solves the Udemy Mix Shift Issue from Consumer to Enterprise

Over the past three years, Udemy management attempted to pivot the company away from an economically challenged consumer transactional marketplace toward a higher margin enterprise subscription business. The company was largely successful in its effort but unfortunately revenue stagnated, even as the quality of revenue and profitability improved.

As enterprise revenue scaled, gross margins expanded and operating losses narrowed, reflecting a fundamentally healthier business mix.

In Q3 2025, Udemy’s CEO provided rare clarity on the consumer segment’s fundamental weakness, acknowledging a Lifetime Value (LTV) to Customer Acquisition Cost (CAC) ratio of 1x.

For those unfamiliar with this metric, LTV/CAC measures the total economic value a customer contributes over their lifetime relative to the cost required to acquire that customer. A ratio of 3x is typically considered the benchmark for a healthy business. At 1x, the company recoups acquisition costs after accounting for content creator revenue share. I don’t know their internal definition of LTV/CAC, so it’s unclear if this calculation includes allocated corporate overhead. It’s conceivable that the company loses money on this part of the business.

This revelation explains why management redirected resources toward enterprise offerings, which now account for nearly 70% of total revenue.

If Udemy had been privately held, management would have had the flexibility to restructure the business and consider extreme measures regarding this potentially unprofitable consumer business. Management had two options. First, to exit the business of transactional consumer purchases and focus solely on selling subscriptions that have a much higher LTV/CAC. Subscription consumer revenue represented 17% of total consumer revenue in 2025. Or second, to exit the consumer business altogether. In the second option, the company would have seen a temporary revenue reset, but would have had a structurally higher-margin enterprise business. Public market optics in my opinion prevented that outcome.

As a result, Udemy continued to subsidize a legacy model with poor unit economics.

New Coursera creates the opportunity Udemy lacked as a standalone public company. Under the umbrella of a strategic transaction, management now can rationalize or eliminate consumer advertising spend for Udemy consumer offerings. This decline in revenue can be reframed by New Coursera management as a synergy to unlock margin expansion that was always economically justified but difficult to execute.

Overall Benefits

Big picture the merger accomplishes the following:

• Optimizes the consumer and enterprise channels as described above

• Eliminates duplicative costs

• Fixes the challenge of shifting Udemy’s business from consumer to enterprise

• Eliminates future investments associated with AI content creation and user tools

• Eliminates future investments associated with integrating the enterprise product into the HR tech stack

Valuation

Based on 2026 consensus estimates and pro forma adjustments, the combined entity is trading at liquidation-level multiples in my opinion.

At pro forma ~1.9x EV/2026 EBITDA, the market is effectively assuming business cessation at some point over the next several years. If New Coursera merely survives, without material revenue growth, it could trade closer to the 6–7x EV/EBITDA range implied by publicly traded education and training peers.

The publicly traded for-profit education companies are also in the business of providing credentials. One key difference is that they own their brands (e.g., Phoenix Education Partners owns University of Phoenix) while New Coursera rents its university and corporate brands non-exclusively. That being said, the brands that Coursera rents may have greater brand appeal than the publicly traded for-profits.

Below I provide my calculations for the implied 1.9x EV/EBITDA multiple for New Coursera on a pro forma basis. I calculate that New Coursera would trade at 1.5x EV/EBITDA with my hypothesis of $200M in expense synergies.

Risks

Here are risks related to the thesis I outlined:

• Merger execution risk. The integration will clearly be complex, involving platform consolidation, go-to-market alignment, cost rationalization, and cultural integration. A lot can go wrong. Managing employee churn is key.

• Management of Udemy content creators. Udemy is a marketplace for content creators to create courses sold to the enterprise and consumers. I imagine that the content creators are nervous about this merger. They should be more nervous about AI affecting the value of their content. The merger is a net positive for them, as a stronger New Coursera protects them from AI substitution.

• AI substitution narrative persists despite beat and raise quarters. Even if fundamentals remain intact, investor sentiment may continue to penalize edtech businesses perceived as vulnerable to AI, suppressing valuation multiple expansion.

• Credential relevance erosion. If employers increasingly on internal skill assessments, internal training, or AI evaluation of skills rather than third-party credentials, the long-term value of external credential could decline.

• Deal may not go through. Deal expected to close in the 2H of 2026. Given the stock price performance, it’s conceivable that a select group of shareholders may seek to terminate the merger agreement.

• New Coursera does not own the credentials that it sells. Rather, it licenses these credentials from third parties like corporations and universities. Management needs to sign long-term contracts with its content partners to provide protection.

Then there are the typical risks associated with any edtech company, like macroeconomic risk, loss of key employees, etc…

Conclusion

New Coursera trades at a distressed valuation multiple. The market clearly is assuming that much will go wrong presumably driven by AI displacement. We haven’t seen any of this yet. Perhaps that’s what Newspaper Industry analysts wrote with the advent of the Internet.

The core question is not whether AI changes the model. It will.

The question is whether credential signaling retains value in a labor market reshaped by automation. If the answer is yes, then I think that optimism is warranted at current price levels.

I’ll end this rather extended article with a quote by Charles Darwin:

“It is not the strongest of the species that survives, nor the most intelligent, but the one most responsive to change.”

The Coursera / Udemy merger is a good step in the right direction.

Appendix - Case Study of Chegg

Chegg once commanded a multibillion dollar market capitalization. Its value plummeted by more than 99% to less than $70M. This collapse was driven entirely by AI substitution of the company’s core product. This outcome is what I suspect many former owners of New Coursera stock feared.

Chegg Study was a subscription based “homework helper”. It offered textbook solutions and expert Q&A to college students. The product was controversial. Critics argued that it helped students cheat rather than fostered genuine learning. Chegg’s product, sold to consumers, lacked institutional validity.

Chegg ultimately sold just content. Not outcomes. The product had no moat. Because Chegg’s value proposition was providing answers, it was immediately replaced by LLMs that could replicate the same output at a lower cost.

Chegg did not provide its learners with a credential. It helped learners progress (some would say cheat) through a course with a path toward receiving a credential at their respective universities.

I’ll submit that any prospective investor should sign up for both Chegg Study and Coursera. You see very different marketing messages to the consumer. Chegg offers “homework solutions”, “quiz & exam help”, “plagiarism detection”, etc…

Meanwhile Coursera offers an actual certificate with a branded entity. In the screenshot below, you can see that Coursera offers an IBM branded product manager professional certificate.

But just visiting the landing page of each company does not do justice to the value of the Coursera course. The output is the digital certification from the branded entity. See below the digital certificate I received from Coursera after taking a course from Google, placed on my LinkedIn account. In a recent Novoresume survey of over 200 HR professionals in the United States, it was reported that LinkedIn has become a meaningful input in hiring, with the vast majority of HR professionals using candidate profiles in their evaluations and a sizable minority viewing them as decisive when identifying top talent.

Disclaimer / Disclosure

The author beneficially owns shares of Coursera, Inc. (NYSE: COUR) and Udemy, Inc. (NASDAQ: UDMY). As a result, the author has a financial interest in the performance of these securities, and readers should consider this potential conflict of interest when evaluating the analysis and conclusions presented herein.

This document reflects the author’s personal views and opinions as of the date of publication and is subject to change without notice. The information contained herein has been obtained from publicly available sources believed to be reliable; however, the author makes no representation or warranty, express or implied, as to the accuracy, completeness, or timeliness of the information.

This material is provided solely for informational and educational purposes and does not constitute investment advice, legal advice, accounting advice, or tax advice. Nothing herein constitutes an offer to sell, a solicitation of an offer to buy, or a recommendation regarding any securities transaction. Any forward-looking statements are inherently subject to risks, uncertainties, and assumptions, and actual results may differ materially.

Investors should conduct their own independent due diligence and consult their financial, legal, and tax advisors before making any investment decisions.

How is artificial intelligence changing industry economics?

AI is already reshaping the labor market, workflows inside companies, and the competitive dynamics of entire industries. For investors, the challenge is separating hype from the real structural changes happening inside businesses.

This newsletter will approach that question through the lens of an investor analyzing publicly traded companies.

In particular, I plan to examine how AI is altering the way companies acquire customers, operate their businesses, and ultimately generate profits.

A natural starting point for me is education technology. Much of my professional background has involved analyzing companies in the education and training ecosystem, which provides a useful vantage point for understanding how AI may affect credentialing, online learning platforms, and workforce reskilling. I expect to begin there and expand into other sectors over time.

A bit about my background.

Earlier in my career I worked on the sell-side at several investment banks analyzing publicly traded companies across industries including software, media, and education technology. That work involved building financial models, evaluating earnings results, studying regulatory filings, and forming investment views about how businesses create long-term shareholder value.

That analytical framework is what I intend to bring to this newsletter.

Beginning in 2025, I started publishing research and commentary on platforms such as Seeking Alpha and LinkedIn. Those platforms have been useful for sharing ideas, but they do not provide much visibility into who is actually reading the work or allow a direct relationship with readers.

Substack changes that dynamic.

It allows writers to build a direct relationship with readers through email subscriptions and provides ownership of the distribution channel.

Stay tuned.