Is the AI in Education Market growing now?
In our AI in education market deck, you will find everything you need to understand the market
SUMMARY
Yes. The AI in Education market is growing now, but adoption and institutional deployment are moving much faster than standalone education-software revenue or venture funding.
Student adoption is already close to saturation in some university populations. The next phase is less about getting students to try AI and more about controlling how deeply it enters coursework, assessment, tutoring and institution-approved learning environments.
Teachers may be the cleaner commercial opportunity. AI is increasingly attached to repetitive, measurable work such as lesson preparation, worksheets, differentiation and administration, which gives schools a much easier return-on-investment argument than broad promises about personalized learning.
The institutional market is becoming more credible because deployments are moving beyond pilots. California State University's decision to renew a system-wide ChatGPT Edu agreement for another three years is stronger evidence of demand than another university announcing an experiment.
AI-native education startups are also finding a recognizable route to market. MagicSchool, Brisk and SchoolAI built large teacher user bases first, then converted part of that bottom-up adoption into school and district relationships around governance, privacy, administration and student access.
Funding tells a much less dramatic story. Wider EdTech venture funding is down year over year, and the latest disclosed pure-play AI education rounds remain relatively small, so the sector looks nothing like a full venture-capital boom yet.
Where investors put the larger checks is revealing. More capital is flowing toward curriculum platforms, practice products and institutional infrastructure than toward generic AI tutors, suggesting that owning a workflow is becoming more valuable than simply wrapping an LLM in an education interface.
Competition from OpenAI, Google and Anthropic is quickly removing basic AI access as a defensible product. Schools and teachers can increasingly get governed or education-specific AI through existing platforms, sometimes at no additional cost.
The learning evidence is promising but uncomfortable for weaker products. Structured AI tutors and human-AI teaching systems can improve outcomes, while unrestricted chatbot access can make students perform better during practice without necessarily helping them learn more.
The market has therefore moved past the question of whether AI will be used in education. The harder commercial question now is who can own the layer around that usage: curriculum context, teacher controls, institutional governance, learning data, workflow integration and measurable outcomes.
Our conclusion is that AI in Education is in an adoption boom with an emerging commercial market. Distribution is already enormous; the companies most likely to capture durable revenue are those that save teachers meaningful time or improve learning in ways that general-purpose chatbots cannot easily reproduce.
This market map, featured in our AI in education market deck, highlights top companies and startups in the AI in education market
Is the AI in Education market really growing now, or is this mostly ChatGPT usage?
Yes. The AI in Education market is growing now, although usage is expanding much faster than standalone education-software revenue.
That distinction controls almost everything else in the article. Millions of students can use ChatGPT, Gemini or Claude for schoolwork without a university buying a dedicated education product. Teachers can do the same for lesson planning. Google can add Gemini to software a district already pays for. A huge increase in AI use therefore does not automatically create a huge new software market.
Still, there is now enough activity beyond casual chatbot use to call this a real market. Schools are buying governed AI tools, universities are signing system-wide contracts, teachers are adopting products built around classroom workflows, and specialized companies such as MagicSchool, Brisk and SchoolAI have converted individual users into school and district relationships.
The financial side is younger. HolonIQ counted only $1 billion of global EdTech venture funding in the first half of 2026, down 26% from $1.35 billion a year earlier. Our review of the latest 12 disclosed pure-play AI education rounds found about $90.4 million combined. Those numbers look small beside the adoption story.
So when we say the AI in Education market is growing, we mean something fairly specific: usage is already massive, institutional deployment is rising, and a commercial layer is forming around it. Revenue and venture funding have yet to catch up with the excitement.
If you want more recent data on this point, please see our latest AI in education market report.
Are students still driving AI in Education growth, or has student adoption already topped out?
Among university students, AI adoption is currently so high that future growth has more to do with deeper use than converting new users.
The clearest longitudinal evidence comes from the Higher Education Policy Institute and Kortext. Their surveys of UK undergraduates found that 66% used AI in at least one way in 2024. That reached 92% in 2025 and 95% in 2026. Use of generative AI for assessed work rose even faster, from 53% to 88% and then 94%.
Look at the shape of that curve. General AI use gained 26 percentage points between the first two surveys, then only another three. Assessed-work use jumped 35 points, then six. AI keeps spreading through student work, but the pool of completely new users has almost disappeared in this population.
The behavior is also becoming more embedded. In the latest HEPI survey, 65% of students said assessment had already changed significantly because of AI. Direct inclusion of AI-generated text in assessed work rose from 3% in 2024 to 8% in 2025 and 12% in 2026.
We should therefore stop treating “students discover AI” as the main growth story in higher education. The bigger opportunities now sit around what happens after adoption: institution-approved access, course-specific assistants, assessment redesign, AI literacy, tutoring and controls over how students use the technology.
| UK undergraduate AI use | 2024 | 2025 | 2026 |
|---|---|---|---|
| Used AI in at least one way | 66% | 92% | 95% |
| Used GenAI for assessed work | 53% | 88% | 94% |
| Included AI-generated text directly in assessed work | 3% | 8% | 12% |
As this chart shows, and as featured in our AI in education market deck, search interest in AI tutors has increased rapidly
Are teachers using AI enough to create a real AI education market?
Yes. Teacher use has become one of the strongest current growth drivers in AI education because it connects AI to repetitive work that schools already spend money and staff time on.
Gallup surveyed 2,232 U.S. public-school teachers and found that 60% had used AI for work during the 2024-25 school year. Nearly one-third, 32%, were already using it every week. The most common recurring jobs were preparing lessons, creating worksheets or activities and adapting material for different student needs.
The time savings make this commercially interesting. Teachers who used AI weekly estimated that it saved them 5.9 hours a week on average. Among teachers already using AI for particular tasks, 84% said it saved time when making worksheets or assignments, 83% when creating assessments, 81% for administrative work and 80% when preparing to teach.
Those are much easier problems to sell against than a vague promise of “personalized education.” A district can understand what it means to cut hours from lesson preparation, differentiation or paperwork.
The market has lately started moving directly toward that opportunity. Anthropic launched Claude for Teachers in the U.S. with free premium access for verified K-12 educators, curriculum resources mapped to standards in all 50 states and tools for lesson planning, differentiation and class-data analysis. The launch also shows that large model providers see teachers as a distinct user group worth building for.
Teacher AI now has a much clearer commercial shape than it did two years ago. The product gets used repeatedly, the pain is obvious, and the buyer can measure whether it saves time.
Are schools and universities actually paying for AI now?
Yes. Schools and universities are currently moving real budgets into AI, and some of the largest deployments have already survived their first renewal test.
California State University is the best example because of its scale. The system originally rolled out ChatGPT Edu to more than 460,000 students and over 63,000 faculty and staff across its campuses. In 2026, CSU renewed the agreement for another three years at $13 million per year, according to reporting based on the university's confirmation. The previous 18-month agreement had cost $17 million.
More importantly, the technology is actually being used inside the system. A CSU survey covering more than 94,000 students, faculty and staff found ChatGPT was used by 84% of responding students, 87% of faculty and 89% of staff. Roughly 30% of students and faculty said they used ChatGPT daily or more often. Faculty adoption had also reached teaching itself: 55% said they used AI to develop course materials and 69% gave students guidance on responsible AI use.
Anthropic has pursued the same institutional market from another angle. Claude for Education launched with campus-wide agreements at Northeastern, the London School of Economics and Champlain College. Northeastern alone made Claude available to 50,000 students, faculty and staff across 13 campuses.
These deployments still face resistance. CSU's renewal triggered faculty criticism over cost, reliability and educational value. But the renewal itself is useful evidence: a large public university system had a chance to walk away after the first deployment and chose another three years instead.
For us, that is a stronger market-growth test than another pilot announcement.
This chart, featured in our AI in education market deck, illustrates yearly VC funding for AI in education startups
Are AI education startups actually turning millions of users into businesses?
A small group of AI education startups is now showing real commercial traction, especially when free teacher adoption becomes a district contract.
MagicSchool is the strongest current example. Recent reporting on the company's scale puts it at roughly eight million registered educators worldwide, with partnerships covering large U.S. districts including Denver Public Schools, Broward County and Hillsborough County. Founder Adeel Khan says about one in five U.S. children attends a school that has partnered with MagicSchool. The company has raised nearly $63 million in total and reported roughly threefold year-over-year revenue growth at the end of last year.
Brisk followed a similar path from teachers into schools. When it raised its $15 million Series A, the company reported one million educators across more than 100 countries, a fivefold increase in users since its previous round and partnerships with more than 2,000 schools.
SchoolAI said it was embedded in more than 400 school districts when it raised a $25 million Series A. The company also reported use across more than one million classrooms in all 50 U.S. states and over 80 countries.
We should treat user and reach figures from private companies with some caution because they are self-reported and usually include free users. Still, the pattern across three separate companies is hard to dismiss. Teachers adopt the product first, schools notice the usage, and the vendor then sells privacy, controls, student access, curriculum alignment and administration at the institutional level.
That bottom-up route is starting to look like the native go-to-market model for K-12 AI.
If you want more recent data on this point, please see our latest AI in education market report.
Is venture funding confirming an AI in Education boom?
No. AI in Education funding is active these days, but investors are still writing relatively modest checks and the wider EdTech funding market remains weak.
HolonIQ counted $1 billion of global EdTech venture capital in the first half of 2026, down 26% year over year. Deal volume stayed roughly stable. Investors have kept making bets while shrinking the amount of capital going into the average company.
The pure-play AI education deals we reviewed show the same thing. The latest 12 publicly disclosed rounds through mid-July raised about $90.4 million combined. Subject's $28 million growth round was the largest. Gizmo raised $22 million, Nectir $12.5 million and Pensive $6.8 million. The remaining rounds were all below $6 million.
There is plenty of company formation here. What we do not see yet is the string of $100 million, $200 million and $500 million rounds that usually appears when venture investors believe a new software category has reached obvious breakout scale.
The financing picture is therefore quite different from the usage picture. AI is already common inside education. Venture capital is still testing which business models deserve large amounts of money.
This chart, featured in our AI in education market deck, breaks down Turnitin’s playbook in AI in education
What are investors actually backing in AI education right now?
Investors currently appear more willing to put larger checks into learning platforms, practice products and institutional infrastructure than into another standalone AI tutor.
We found five clear tutoring companies among the latest 12 disclosed pure-play AI education rounds: sortmyprep, YoLearn.ai, ProLearn, Third Space Learning and Lucida AI. Together they raised about $15.4 million, only 17% of the $90.4 million total.
Three other companies alone absorbed $62.5 million. Subject raised $28 million for an AI-powered K-12 curriculum and online-learning platform, Gizmo raised $22 million around AI study and practice tools, and Nectir raised $12.5 million for higher-education AI infrastructure and course-specific assistants. Those three rounds represented roughly 69% of all capital in the 12-deal sample.
That concentration tells us more than simply counting how many “AI tutor” startups exist. Investors are still willing to fund tutoring experiments, but the larger checks are going to products that control more of the learning workflow or plug directly into an institution.
A generic tutor has a brutal benchmark these days: ChatGPT, Gemini and Claude can already explain algebra, revise an essay or quiz a student. A curriculum platform, governed university assistant or deeply embedded practice product has more room to own a workflow that general chatbots do not automatically control.
If you want more recent data on this point, please see our latest AI in education market report.
Are OpenAI, Google and Anthropic making life harder for AI education startups?
Absolutely. General-purpose AI companies are moving directly into education now, and free or heavily subsidized products are raising the bar for every specialized vendor.
Google already includes Gemini for Education as a core Workspace service for qualifying education customers at no additional charge. Schools get an admin-managed environment and enterprise data protections without buying another standalone chatbot.
Anthropic has gone further into the teacher workflow with Claude for Teachers, giving verified U.S. K-12 educators premium capabilities for free. The product connects to curriculum resources and academic standards, which starts to overlap directly with features sold by teacher-focused startups.
OpenAI has recently pushed further into the student side. Study Mode is available across ChatGPT plans and is designed around guided questions, step-by-step explanations and checks for understanding. ChatGPT for Teens is also rolling out with Study Mode, homework reminders and learning-focused flows for users aged 13 to 17.
Pricing shows the pressure clearly. UC Davis renewed ChatGPT Edu for faculty and staff at $144 per user per year, down from $240, a 40% cut. Specialized vendors therefore face competitors that are simultaneously becoming more education-specific and cheaper.
| Education AI offer | Current pricing signal | Why it matters |
|---|---|---|
| Gemini for Education | Included with qualifying Workspace for Education editions | Schools can get governed AI without adding a new chatbot vendor |
| Claude for Teachers | Free for verified U.S. K-12 teachers during the current offer | Direct pressure on teacher copilot products |
| ChatGPT Study Mode / Teens | Available through existing ChatGPT access | General ChatGPT is becoming a more credible learning product |
| ChatGPT Edu at UC Davis | $144 per user annually, down 40% | Institutional AI pricing is already compressing |
This competitive pressure will probably kill weak wrappers quickly. The surviving education companies will need to win through workflow, curriculum, data, controls or measurable outcomes rather than basic access to a good model.
This chart, featured in our AI in education market deck, illustrates yearly funding for AI in education startups
Do AI tutors actually help students learn?
Sometimes, and the latest evidence is much less magical than the early AI-tutor narrative suggested.
A newly released NBER study gives us one of the hardest tests so far. Researchers ran a two-year cluster-randomized experiment across 18 Tennessee middle schools using Khan Academy with Khanmigo during daily remedial mathematics sessions. Assignment to the program raised math achievement by about 1.3 national percentile ranks per term, equivalent to roughly 0.06 to 0.08 standard deviations over a school year.
That is a real effect. Yet the researchers found that the gains looked similar to what Khan Academy practice had achieved without AI assistance. Students used Khanmigo relatively infrequently, and the estimated effect for a full year of active participation was around 0.14 standard deviations.
Earlier controlled experiments produced stronger results under tighter conditions. A Harvard study of 194 physics students found that students using a carefully designed AI tutor achieved more than twice the median learning gains of students in an active-learning classroom condition, while typically spending less time on the lesson.
A high-school mathematics experiment involving nearly 1,000 students showed why we should be careful. Students with access to a standard GPT-4 interface performed 48% better during practice, while a guarded GPT Tutor produced a 127% improvement. When access disappeared for the later exam, students from the unrestricted GPT group performed 17% worse than students who had never received AI access. The guarded version largely removed that damage.
So the evidence today supports a narrower claim than “AI tutors work.” AI can improve learning when the product makes students think, gives the right kind of help and gets used enough. A chatbot that simply makes homework easier can produce impressive practice scores while teaching very little.
If you want more recent data on this point, please see our latest AI in education market report.
Does purpose-built education AI have an advantage over a normal chatbot?
Yes. Purpose-built education AI has a real advantage when it controls how the model teaches, what information it uses and what the teacher can see.
The high-school math experiment above gives us direct evidence. Two interfaces built on the same underlying GPT-4 capability led to very different learning outcomes because one was designed to prevent students from turning the model into an answer machine.
Tutor CoPilot gives us another example from the teacher side. In a randomized study involving roughly 900 tutors and 1,800 K-12 students, students whose tutors had access to AI assistance were four percentage points more likely to master lesson topics. The improvement reached nine percentage points for students working with lower-rated tutors. The researchers estimated the AI cost at only about $20 per tutor per year during the study.
The product design there is revealing. Tutor CoPilot suggests teaching strategies in real time, such as asking a guiding question, giving a hint or using a similar problem. The human tutor still decides what to say.
The successful commercial products are moving in the same direction. SchoolAI gives teachers visibility over student interactions. Nectir builds assistants around instructor-approved course material. MagicSchool combines teacher workflows with managed student experiences.
We would be much more skeptical of a startup whose main feature is “ChatGPT with an education prompt.” The useful vertical layer is becoming deeper: curriculum context, teacher control, student safeguards, institutional privacy, learning data and behavior designed around how people actually learn.
This chart, featured in our AI in education market deck, compares the main business model options for AI tutoring platforms
What is still slowing the AI in Education market down?
Schools currently have a bigger governance and proof problem than an awareness problem.
Gallup's latest U.S. teacher research makes the gap obvious. Only 18% of K-12 teachers said they received formal guidance from administrators on how AI should be used at work. For one-on-one instruction or tutoring, 69% said they had received no guidance at all. For grading and feedback, 58% said the same.
That sits awkwardly beside the 60% teacher adoption rate we saw earlier. Teachers are already using AI while many schools are still figuring out the rules.
The evidence around learning also remains mixed enough to slow large purchases. The OECD's 2026 Digital Education Outlook warns about “false mastery”: AI can improve the quality of a student's output without building the underlying skill. The new Khanmigo experiment makes the same problem more concrete from another angle. Even a purpose-built AI tutor can deliver only modest incremental gains if students barely engage with the tutoring layer.
Quality is another practical issue. A recent study tested 11 general and education-specific AI tools on their ability to classify the cognitive demand of mathematics tasks. Average accuracy was only 63%. That is hardly disastrous for an assistant whose work a teacher reviews, but it is uncomfortable if schools expect the system to make autonomous instructional decisions.
These constraints will probably slow procurement more than usage. Students and teachers can experiment immediately. A school district has to think about student privacy, age restrictions, hallucinations, curriculum quality, teacher training and whether the product actually improves learning enough to justify another contract.
Is the AI in Education market growing now?
Yes. The AI in Education market is clearly growing now, but we would call it an adoption boom with an emerging commercial market rather than a full venture and revenue boom.
Student AI use has already reached near-saturation in some higher-education populations. Teacher adoption has moved into the mainstream. Large university systems are renewing AI contracts instead of ending the experiment. AI-native education companies have reached millions of educators and hundreds or thousands of institutional customers. Purpose-built tutoring and teacher-assistance products now have controlled studies showing that carefully designed AI can improve learning or teaching quality.
The money is moving more slowly. Global EdTech funding is currently down year over year, our latest 12 pure-play AI education rounds total only about $90 million, and most are still small. At the same time, Google, OpenAI and Anthropic are pushing education-specific capabilities into products that are free, bundled or getting cheaper.
AI has already won distribution in education. Companies are now fighting over who gets paid for the layer around that usage.
For specialized vendors, the easy part of the market is disappearing fast. Simply offering students or teachers access to an LLM has little value anymore. The companies with the best chance today are the ones that own a real education workflow, fit inside institutional rules and can show that their product saves teachers time or helps students learn better.
Our final judgment is mostly true, with high confidence: the AI in Education market is growing now. The strongest growth is happening in teacher tools, governed institutional AI, curriculum-linked assistants and structured learning products. Funding data still looks too restrained for us to call the sector a financial boom, but the underlying adoption has moved far beyond experimentation.
If you want more recent data on this point, please see our latest AI in education market report.
This chart, featured in our AI in education market deck, illustrates how market revenue is distributed across customer segments in the AI in education market
OUR METHODOLOGY
This analysis tests whether the AI in Education market is growing based on evidence of actual usage, institutional purchasing, startup traction, venture funding, competitive pressure and educational outcomes. We do not treat a rise in general ChatGPT, Gemini or Claude usage as proof by itself that a standalone education software market is expanding.
We assess student and teacher adoption separately from commercialization. High usage tells us that AI already has distribution inside education, while university contracts, school and district relationships, pricing and startup revenue give us a better view of who is actually getting paid.
For student adoption, we use the Higher Education Policy Institute and Kortext surveys to track how undergraduate AI use has changed over several years. For teachers, we rely primarily on Gallup's U.S. public-school teacher research, including frequency of use, common workflows, guidance from administrators and reported time savings.
Institutional deployments receive more weight when there is evidence of commitment beyond an initial pilot. California State University's system-wide ChatGPT Edu deployment and subsequent three-year renewal are particularly useful because they combine scale, reported usage and a repeat purchasing decision. We also use Anthropic's university agreements and UC Davis pricing as additional evidence of how institutional AI purchasing is developing.
For startup traction, we distinguish self-reported user reach from institutional relationships. Large free user counts are useful for showing distribution, but school and district partnerships are more relevant to commercialization. MagicSchool, Brisk, SchoolAI and Nectir are used as examples of companies trying to turn bottom-up adoption into paid institutional products.
Our funding analysis looks at both the wider EdTech financing environment and the latest publicly disclosed pure-play AI education rounds in our sample. We use the distribution of capital across tutoring, curriculum, study products and institutional infrastructure to understand which business models are attracting the larger checks, rather than relying only on the number of startups being funded.
We also account for competition from general-purpose AI providers because it changes what counts as a defensible education product. Google, Anthropic and OpenAI increasingly offer education-specific features, governed access or teacher and student workflows inside broader AI platforms, sometimes free or bundled into products institutions already use.
For educational outcomes, we prioritize controlled studies over product claims. The NBER Khanmigo experiment, the high-school mathematics study published in the Proceedings of the National Academy of Sciences, the Harvard physics experiment and the Tutor CoPilot trial help separate better AI-assisted performance from actual learning and show how much product design can change the result.
Key sources used for this analysis include: HEPI's Student Generative Artificial Intelligence Survey 2026, Gallup's research on teacher AI use and time savings, California State University information on its ChatGPT Edu contract extension, CalMatters on the CSU deployment, renewal and institutional debate, Anthropic on Claude for Teachers, Anthropic on Claude for Education, Google on Gemini for Education, OpenAI on ChatGPT Study Mode, UC Davis on its renewed ChatGPT Edu agreement and pricing, NBER's two-year Khanmigo experiment, the PNAS study on generative AI and high-school mathematics learning, the Tutor CoPilot randomized trial, and the OECD Digital Education Outlook 2026.
This chart, featured in our AI in education market deck, shows how AI conversational tutor technology has evolved over time
Related blog posts
