185 AI in Education Statistics October 2026

AI in education statistics for 2026 show students moving faster than the schools around them. 84% of US high school students used generative AI for schoolwork by May 2025 (College Board, Oct 2025), yet over 80% said their teachers had not explicitly taught them how to use it (RAND, Sep 2025).

The page covers student use, teachers and faculty, school and government rules, cheating and detectors, learning outcomes from controlled studies, and the money and access gaps behind all of it. Each one links to the organisation that produced it: governments and regulators, peer-reviewed journals, large surveys, and vendors and market-research firms, which are labelled as such. For the wider picture, see our AI statistics, generative AI statistics, AI usage statistics and ChatGPT user statistics pages.

How many students use AI for school, and what for?

Students adopted first and most were never taught how. Use is close to universal at university and a majority in secondary school, and most of it is about explaining, summarising and drafting.

How many students use AI for schoolwork?

94% of UK undergraduates used generative AI to help with assessed work in the 2025-26 academic year (HEPI, Mar 2026).

• 86% of students in a survey of more than 3,800 students in 16 countries used AI in their studies, and 54% of those used it weekly (Digital Education Council, Aug 2024).
• 66% of those students used ChatGPT, the most-used tool in the survey (Digital Education Council, Aug 2024).
• 42% of students in US introductory college courses used generative AI weekly in 2025, against 30% of instructors (Tyton Partners, Jun 2025).
• 54% of US students said they used AI for school in winter 2025, and 61% of high school students did (RAND, Sep 2025).
• 69% of US high school students used ChatGPT for school assignments and homework in May 2025 (College Board, Oct 2025).
• Nearly 60% of high school students used generative AI often or sometimes for schoolwork, across more than 10,000 student, parent and teacher responses (College Board, Dec 2025).
• 54% of US teens used chatbots to get help with schoolwork, and 35% used them to edit something they wrote (Pew Research Center, Feb 2026).
• 40% of US teens had used generative AI for school assignments by spring 2024, and for their latest assignment 46% used it without the teacher's permission against 41% with it (Common Sense Media, Sep 2024).

The surveys use different ages, countries and definitions, so the levels differ. Every wave shows growth, and the growth is fastest where schools have said least.

What do students use AI for?

Text generation is still the top use for UK undergraduates at 56%, down from 64% a year earlier, as students move to narrower tools (HEPI, Mar 2026).

• 38% used AI to summarise, take notes or quiz themselves on textbooks, and 37% used it to enhance and edit their writing (HEPI, Mar 2026).
• 31% used it for translation or language support, 22% to generate images, video or audio, and 17% to write code (HEPI, Mar 2026).
• Half of US high school students used AI to brainstorm ideas, edit or revise essays, or research and find sources (College Board, Oct 2025).
• 15% of UK undergraduates used AI for friendship, company, advice or loneliness, and 8% used counselling or therapy provided solely by AI (HEPI, Mar 2026).
• 49% of UK undergraduates said AI had improved their student experience, mainly by saving time, improving understanding and giving instant support (HEPI, Mar 2026).
• About 47% of 574,740 anonymised university-student conversations with Claude were "direct", meaning students sought answers with minimal engagement (Anthropic, Apr 2025). This is vendor data from one product.
• Business accounted for 8.9% of those conversations against 18.6% of US bachelor's degrees, health professions for 5.5% against 13.1%, and humanities for 6.4% against 12.5% (Anthropic, Apr 2025).
• 42% of UK undergraduates agreed AI content could earn a good grade in their subject, but only 25% in arts and humanities (HEPI, Mar 2026).

Students mostly use AI to understand and to draft. A large minority use it to get the answer and stop there.

Do students think AI helps or hurts their learning?

Among students who had met AI in their courses, only 5% said it transformed how they learn and 28% said it improved their understanding and outcomes (Digital Education Council, Jun 2026).

• 42% of those students said AI integration was only somewhat helpful, and 24% saw no clear learning value (Digital Education Council, Jun 2026).
• Only 15% of students said AI was built into many of their courses, 43% said a few and 43% said none (Digital Education Council, Jun 2026).
• 55% of US high schoolers, 48% of middle schoolers and 61% of parents agreed that more AI use will harm students' critical thinking, against 22% of district leaders (RAND, Sep 2025).
• 50% of US high school students neither agreed nor disagreed that AI's benefits outweigh its risks, and a quarter fell on each side (College Board, Oct 2025).
• 68% of UK undergraduates said AI skills are essential, but 48% felt their teaching staff were helping them build those skills (HEPI, Mar 2026).
• 58% of students in 2024 said they lacked sufficient AI knowledge and skills, and 48% felt unprepared for an AI-enabled workplace (Digital Education Council, Aug 2024).
• 80% of students said AI at their university was not fully meeting expectations, and only 5% were fully aware of their institution's AI guidelines and thought them fully comprehensive (Digital Education Council, Aug 2024).
• 41% of students worried AI would reduce job opportunities in their field by graduation, rising to 50% in Asia Pacific (Digital Education Council, Jun 2026).
• Only 28% of students felt most of their assessments reflected the skills and judgement an AI-enabled workplace would ask of them (Digital Education Council, Jun 2026).

Students want AI skills and doubt that their courses supply them. Their own worry about weaker thinking is higher than the worry of the people who run districts.

How many teachers and professors use AI, and who trains them?

Teachers took up AI to save time, and most taught themselves. Training reaches a minority and professors are less confident than their usage suggests.

How many teachers use AI?

Across OECD systems, 36% of lower secondary teachers had used AI in their work in the 12 months before the 2024 TALIS survey (OECD, Jan 2026).

• TALIS surveyed 280,000 educators in 55 education systems, which makes it the largest survey of teachers and principals (OECD, Oct 2025).
• 60% of US K-12 teachers used an AI tool for work in 2024-25, 32% at least weekly and 28% monthly or less (Gallup, Jun 2025).
• Use reached 66% among high school teachers and 69% among early-career teachers (Gallup, Jun 2025).
• About 13% of US English, math and science teachers used AI at least once a week, and use rose with grade level from 42% in elementary school to 64% in middle school and 69% in high school (RAND, Sep 2025).
• In England, 50% of teachers had used generative AI in their role in November 2024 (UK Department for Education, Nov 2024).
• That share was 82% of primary teachers and 78% of secondary teachers in December 2025 (UK Department for Education, Dec 2025).
• A fall 2023 survey of 2,531 US public K-12 teachers found 25% saying AI tools do more harm than good, 6% more good than harm, 32% an equal mix and 35% unsure (Pew Research Center, May 2024).
• 35% of high school teachers said AI did more harm than good, against 24% in middle school and 19% in elementary school (Pew Research Center, May 2024).

Teacher use rose sharply in a year or two in the US and England, and the OECD data show the level depends heavily on the country.

What do teachers use AI for, and how much time does it save?

Teachers who used generative AI in England most often created resources, at 84% of secondary and 75% of primary users (UK Department for Education, Dec 2025).

• 61% of primary and 54% of secondary users used it to plan lessons, and 53% and 47% to adapt materials to individual pupils (UK Department for Education, Dec 2025).
• 45% of primary users used it to support pupils with special educational needs, against 29% of secondary users (UK Department for Education, Dec 2025).
• 34% of secondary users used it for assessments, plagiarism checks, marking or feedback, against 8% of primary users (UK Department for Education, Dec 2025).
• Only 15% of primary and 14% of secondary users used it during live teaching (UK Department for Education, Dec 2025).
• US teachers used AI at least monthly to prepare to teach (37%), make worksheets or activities (33%) and modify materials to student needs (28%) (Gallup, Jun 2025).
• 60% to 84% of teachers said AI saved them time on each of nine tasks, and 7% or fewer said it made the task longer (Gallup, Jun 2025).
• 57% of users said AI improved the quality of grading and feedback and 74% said it improved administrative work, while 16% or fewer reported lower quality (Gallup, Jun 2025).
• Schools with an AI policy averaged 2.3 hours saved per teacher per week, against 1.7 hours in schools without one (Gallup, Jun 2025).
• Among TALIS teachers who used AI, 68% used it to learn about and summarise topics, 64% to generate lesson plans, 25% to review student participation or performance data and 26% to assess or grade student work (OECD, Jan 2026).
• 52% of teachers agreed AI helps them adapt material to different abilities, 45% that it automates administrative tasks and 40% that it helps them support students individually (OECD, Jan 2026).
• 66% worried about incorrect or inappropriate AI recommendations and 42% about AI amplifying biases (OECD, Jan 2026).

The pattern is consistent across countries: drafting, adapting and admin first, grading and live teaching last.

Are teachers and professors trained to use AI?

Mostly not. 68% of US teachers had no school or district AI training in 2024-25, and 52% taught themselves against 31% who were trained by their school or district (Gallup, Jun 2025).

• 55% of US teachers said their school or district provided professional development or resources on generative AI in fall 2024, 65% in high school and 47% in elementary school (RAND, Sep 2025).
• Only 35% of the teachers who got that training found it somewhat or very helpful (RAND, Sep 2025).
• Three in four TALIS teachers who had not used AI said they lacked the knowledge or skills to teach with it (OECD, Jan 2026).
• In England, 61% of primary and 47% of secondary teachers who had not used generative AI said they did not know enough about how to use it (UK Department for Education, Dec 2025).
• 40% of faculty in a 1,681-person survey across 28 countries were just beginning their AI literacy journey and 17% were at advanced or expert level (Digital Education Council, Jan 2025).
• Only 6% of those faculty fully agreed their institution gave sufficient resources to train them in AI, and only 4% were fully aware of their institution's AI guidelines and thought them fully comprehensive (Digital Education Council, Jan 2025).
• 83% of faculty worried about students' ability to critically evaluate AI output (Digital Education Council, Jan 2025).
• 64% of faculty in the 2026 survey had taken part in AI literacy training, yet only 29% of students thought their instructors were well equipped to guide AI use, and 17% in the US and Canada (Digital Education Council, Jun 2026).
• Among more than 3,000 US college faculty surveyed in summer 2025, 77% had used AI in their professional role but only 21% felt very confident guiding AI use in class (College Board, Feb 2026).
• 72% of those faculty faced at least minor challenges managing student AI use, and almost half had a formal classroom AI policy (College Board, Feb 2026).
• UK undergraduates who said staff were well equipped to work with AI rose from 18% in 2024 to 42% in 2025 (HEPI, Feb 2025).
• US and Canadian faculty who intended to use AI in teaching fell from 76% in 2025 to 67% in 2026, and only 26% were excited about AI against 57% in Asia Pacific (Digital Education Council, Jun 2026).

Training lags use almost everywhere. The one rising number is UK students' confidence in their staff.

Who sets the rules for AI in schools, and how clear are they?

Rules trail use by about a school year. Where policies exist they are mostly limited, and governments are writing guidance faster than schools can apply it.

How many schools and universities have clear AI rules?

45% of US principals reported a school or district policy or guidance on AI, and 34% of teachers reported one on academic integrity (RAND, Sep 2025).

• Integrity policies were more than twice as common in high schools (49%) as in elementary schools (22%) (RAND, Sep 2025).
• Only 6% of US teachers called their integrity policies clear and comprehensive (RAND, Sep 2025).
• 37% of teachers who had AI training found their school's AI policies helpful, against 13% of teachers with no training (RAND, Sep 2025).
• Roughly 2 in 5 US high schools or districts did not allow students to use generative AI, and about 1 in 5 allowed it with no policy in place (College Board, Oct 2025).
• 13% of US high schools encouraged generative AI in all classes, 16% left the decision to individual teachers and 12% to departments or grade levels (College Board, Oct 2025).
• 55% of US principals had not blocked students or teachers from generative AI tools on the school network and 45% reported some restriction (College Board, Oct 2025).
• One in ten TALIS teachers said their school prohibits AI use in teaching (OECD, Jan 2026).
• In a 2023 UNESCO survey of more than 450 schools and universities, fewer than 10% had formal guidance on generative AI (UNESCO, Jun 2023).
• 13% of universities had guidance against 7% of schools, and about 40% of institutions with guidance said it was not written down (UNESCO, Jun 2023).
• 37% of UK undergraduates agreed their institution encourages AI use and 36% disagreed (HEPI, Mar 2026).

Counts differ by who is asked and how policy is defined: principals report more policy than teachers do.

Are students allowed to use AI, and who tells them how?

In England, 72% of primary and 71% of secondary teachers said pupils were not permitted to use generative AI in set work, and 3% and 19% said they were (UK Department for Education, Dec 2025).

• 43% of secondary teachers said pupils had been given guidance on using AI safely, against 28% of primary teachers, while 23% of secondary teachers said they had not and 34% did not know (UK Department for Education, Dec 2025).
• 61% of secondary teachers had explained the risks of plagiarism with generative AI to pupils, against 12% of primary teachers (UK Department for Education, Dec 2025).
• Only 9% of secondary and 4% of primary teachers had let pupils use generative AI in class (UK Department for Education, Dec 2025).
• 19% of US students said their teachers had given them guidance on using AI for schoolwork (RAND, Sep 2025).
• 26% of US students said teachers definitely used tools to check whether students used AI, and 40% were unsure (RAND, Sep 2025).
• 35% of US teens said their school had rules on generative AI, 27% said it had none and 37% were not sure (Common Sense Media, Sep 2024).
• 42% of US teens said their teachers mostly did not allow generative AI, and 83% of parents said the school had not communicated with them about it (Common Sense Media, Sep 2024).

Guidance is more common in secondary school, but many teachers cannot say whether pupils received any. In the US, most teens cannot say what the rule is.

What are governments and regulators doing?

30 US states had issued guidance on AI in education as of January 2026 (Stanford HAI, Apr 2026).

• 17 states had clarified computer science as foundational to AI, and five had allocated specific professional development funding for AI education (Stanford HAI, Apr 2026).
• 45 states had adopted K-12 computer science standards, but only six included significant AI content (Stanford HAI, Apr 2026).
• Executive Order 14277 (April 2025) gave the Secretary of Education 90 days to issue guidance on using grant funds for AI instruction and 120 days to prioritise AI in teacher-training grants (The White House, Apr 2025).
• The US Department of Education published its final supplemental priority on advancing AI in education on 13 April 2026, effective 13 May 2026, after more than 300 parties commented on the proposal (Federal Register, Apr 2026).
• England's product safety standards for generative AI in schools, first published in January 2025, gained four new standards in January 2026: cognitive development, emotional and social development, mental health, and manipulation (UK Department for Education, Jan 2026).
• Beijing, Guangdong and Hangzhou required AI education in the 2025-26 school year, and the UAE mandated it for all grade levels (Stanford HAI, Apr 2026).
• South Korea launched AI textbooks in primary schools in March 2025 and reversed course a few months later after parent and teacher pushback (Stanford HAI, Apr 2026).
• Estonia's AI Leap programme is piloting AI learning apps with 20,000 students and 3,000 teachers in 2025-26 (Stanford HAI, Apr 2026).
• State-level AI guidance is largely nonbinding, and the AP Computer Science course has no AI-specific content (Stanford HAI, Apr 2026).

Governments agree that AI literacy matters and disagree on mandates. Most US guidance is advice, and local districts carry the cost of applying it.

What does EU law say about AI in schools?

The EU AI Act lists four education uses as high risk: deciding admission or placement, evaluating learning outcomes, assessing the level of education a person can access, and monitoring prohibited behaviour during tests (EUR-Lex, Jul 2024).

• It bans AI systems that infer the emotions of people in education institutions, except for medical or safety reasons, from 2 February 2025 (EUR-Lex, Jul 2024).
• Providers and deployers must also take measures to ensure a sufficient level of AI literacy among their staff, an obligation that applies from February 2025 (EUR-Lex, Jul 2024).

Tools that proctor tests or evaluate learning outcomes fall on the high-risk list. A tutoring chatbot that does neither is not named there.

Is AI cheating rising, and can anything detect it?

Cheating is rising in the cases universities can prove, but no detector is reliable enough to prove it alone. The errors do not fall evenly.

How common is cheating with AI?

The Guardian asked 155 UK universities for misconduct figures under freedom of information law and 131 provided data (The Guardian, Jun 2025).

• Early figures pointed to about 7.5 proven AI cheating cases per 1,000 students in 2024-25 (The Guardian, Jun 2025).
• Proven traditional plagiarism fell from 19 per 1,000 students to 15.2 in 2023-24, and early figures pointed to about 8.5 in 2024-25 (The Guardian, Jun 2025).
• More than 27% of responding universities did not record AI misuse as a separate category in 2023-24 (The Guardian, Jun 2025).
• In Stanford-affiliated Challenge Success surveys, 60% to 70% of students reported at least one cheating behaviour in the past month, a share that stayed flat or fell slightly after ChatGPT arrived (Stanford Graduate School of Education, Oct 2023).
• 65% of UK undergraduates said assessment had changed significantly in response to AI (HEPI, Mar 2026).
• 74% of US college faculty said students use AI to write essays or papers and 67% said students use it to paraphrase or rewrite content (College Board, Feb 2026).
• 92% of faculty were concerned about AI-enabled plagiarism or dishonesty, and 88% about overreliance on automation (College Board, Feb 2026).
• 72% of TALIS teachers believed AI could let students pass off others' work as their own (OECD, Jan 2026).
• 76% of US teens who had used chatbots for schoolwork said students at their school use chatbots to cheat at least sometimes (Pew Research Center, Feb 2026).
• 77% of US parents said whether AI use for schoolwork counts as cheating depends, 7% said never and almost one in five said always (RAND, Sep 2025).
• 60% of students worried classmates would misuse AI for unfair advantage, rising to 73% in the US and Canada (Digital Education Council, Jun 2026).
• 54% of faculty said current student evaluation methods were no longer adequate in the age of AI (Digital Education Council, Jan 2025).

Recorded cases are a floor. They count only what staff could prove, and the survey shows the proof is hard to get.

How accurate are AI detectors?

OpenAI shut down its own AI-text classifier on 20 July 2023 because of low accuracy: it caught 26% of AI-written text and wrongly flagged 9% of human text (OpenAI, Jan 2023).

• A test of 14 detection tools found all scored below 80% accuracy and only five scored above 70% (International Journal for Educational Integrity, Weber-Wulff et al., Dec 2023).
• Six of the 14 tools produced false positives on human-written text (International Journal for Educational Integrity, Weber-Wulff et al., Dec 2023).
• About half of AI-generated texts that were manually edited or machine-paraphrased would likely be misattributed to humans (International Journal for Educational Integrity, Weber-Wulff et al., Dec 2023).
• In a University of Reading test, 97% of AI submissions went undetected when a flag had to mention AI specifically, and AI answers earned higher median grades than real students in four of five modules (PLOS ONE, Scarfe et al., Jun 2024).
• Not one of the AI-written essay exam answers was detected in that study, although a similarity checker was available (PLOS ONE, Scarfe et al., Jun 2024).
• Turnitin's AI detector flagged 91% of 22 evasion-prompted ChatGPT submissions as containing some AI, but identified only 54.8% of the content as AI-written (Journal of Academic Ethics, Perkins et al., Oct 2023).
• The 15 academic staff who marked those submissions reported only 54.5% of them to the misconduct process (Journal of Academic Ethics, Perkins et al., Oct 2023).
• Turnitin says its document-level false positive rate is below 1% for documents with 20% or more AI writing, and its sentence-level rate is around 4% (Turnitin, Jun 2023). This is the vendor's own figure.
• Vanderbilt University worked out that a 1% false positive rate on the 75,000 papers it submitted in 2022 would have meant about 750 wrongly flagged papers, and switched the detector off (Vanderbilt University, Aug 2023).

A detector gives a probability and proves nothing on its own. A teacher who treats the score as proof is the one making the accusation.

Who gets wrongly accused of using AI?

10% of US teens said a teacher had flagged their work as AI-generated when it was not, and about 70% of those teens said the work was also run through an AI detector (Common Sense Media, Sep 2024).

• 10% of Latino teens reported a false flag, and 29% of teens whose teachers had discussed AI rules said at least one teacher used detection software (Common Sense Media, Sep 2024).
• Seven widely used detectors flagged an average of 61.3% of 91 TOEFL essays as AI-generated while classifying US eighth-grade essays almost perfectly (Patterns, Liang et al., Jul 2023).
• All seven detectors unanimously flagged 19.8% of the human-written TOEFL essays, and at least one flagged 97.8% (Patterns, Liang et al., Jul 2023).
• Prompting ChatGPT to enrich the word choice of the TOEFL essays cut the average false positive rate from 61.3% to 11.6% (Patterns, Liang et al., Jul 2023).
• 51% of US students worried they might be falsely accused of cheating with AI, and 16% were worried because they or someone they knew had been (RAND, Sep 2025).
• 53% of UK students said the risk of being accused of cheating put them off using AI in 2025 (HEPI, Feb 2025).

The same detectors that miss edited AI text flag plain human writing. Writers with a plainer vocabulary, including non-native speakers, pay for it.

Does AI improve learning, or only performance?

The evidence splits cleanly by design. AI tutors built on teaching principles improve learning, and open chatbot access lifts performance while the tool is there and can hurt it once it is taken away.

Do AI tutors beat conventional teaching in controlled trials?

In a crossover trial in Harvard's largest physics class (233 students), the AI-tutored group's median post-test score was 4.5 against 3.5 for the in-class active learning group (Scientific Reports, Kestin et al., Jun 2025).

• A quantile analysis put the effect between 0.73 and 1.3 standard deviations, and 70% of the AI group spent less than the 60 minutes the in-class lesson took (Scientific Reports, Kestin et al., Jun 2025).
• A World Bank randomised trial in Nigeria found a 0.31 standard deviation gain in English after six weeks of GPT-4 tutoring through Microsoft Copilot for first-year senior secondary students (World Bank, May 2025).
• A Ghana trial of the WhatsApp maths tutor Rori covered about 1,000 students in grades 3 to 9 across 11 schools and found an effect size of 0.37 after eight months of two 30-minute sessions a week (arXiv, Henkel et al., Feb 2024). This is a preprint.
• In a randomised trial of Tutor CoPilot with 900 tutors and 1,800 K-12 students, students were 4 percentage points more likely to master topics, and 9 points more for students of lower-rated tutors (arXiv, Wang et al., Oct 2024). This is also a preprint.
• Tutor CoPilot cost about $20 per tutor per year (arXiv, Wang et al., Oct 2024).

The winning designs share a feature: they were built to guide the student instead of answering for them.

Does open access to a chatbot hurt learning?

In a Turkish high school field experiment with nearly 1,000 students, access to a standard GPT-4 interface raised practice scores by 48% and a guardrailed GPT Tutor raised them by 127% (PNAS, Bastani et al., Jun 2025).

• When access was removed, students who had used the standard interface scored 17% lower on the exam than students who never had it (PNAS, Bastani et al., Jun 2025).
• The guardrailed tutor essentially eliminated that loss, though it did not produce a measurable gain over the control group (PNAS, Bastani et al., Jun 2025).
• In a lab experiment with 117 university students, ChatGPT users improved short-term task performance but showed no better knowledge gain or transfer, a pattern the authors call metacognitive laziness (British Journal of Educational Technology, Fan et al., Dec 2024).
• The OECD reports that students with general-purpose chatbots produce higher-quality outputs than peers, but the advantage disappears and sometimes reverses in exams once access is removed (OECD, Jan 2026).
• Among teens who expect AI to have a negative effect on society over 20 years, 34% mention overreliance or a loss of critical thinking or creativity (Pew Research Center, Feb 2026).
• More than 84% of college faculty agreed AI reduces students' critical thinking, originality and engagement with course material (College Board, Feb 2026).

Performance with AI and learning without it are different measurements. Policy that only measures the first will overstate the benefit.

What do reviews of the evidence conclude?

A meta-analysis of 51 experimental studies published between November 2022 and February 2025 found a large positive effect of ChatGPT on learning performance, with Hedges' g of 0.867 (Humanities and Social Sciences Communications, Wang and Fan, May 2025).

• The same review found smaller effects on learning perception (g = 0.456) and higher-order thinking (g = 0.457) (Humanities and Social Sciences Communications, Wang and Fan, May 2025).
• It recommended using ChatGPT for 4 to 8 weeks for more stable effects, and found that the effect varied by course type and learning model (Humanities and Social Sciences Communications, Wang and Fan, May 2025).
• The OECD concludes that educational AI tools designed or used with an intentional pedagogical purpose tend to show sustained improvements in learning (OECD, Jan 2026).
• Students taught by the structured AI tutor in the Harvard trial also reported feeling more engaged and more motivated (Scientific Reports, Kestin et al., Jun 2025).

Short, structured, tutor-style uses show reliable gains, while the long-run effect of unrestricted use is still unmeasured.

How much money flows into AI learning, and who is left behind?

Money has shrunk even as AI use has grown, and the market-size forecasts disagree by a factor of seven. The students who benefit least are often the ones who already had less.

How big is the AI learning market, and why do forecasts disagree?

Grand View Research valued the AI in education market at USD 8.3 billion in 2025, rising to USD 57.2 billion by 2033 at a 25.9% compound annual growth rate (Grand View Research, May 2026).

• Grand View Research put North America's share at 37.5% in 2025 and the 2026 value at USD 11.4 billion (Grand View Research, May 2026).
• Precedence Research estimated USD 9.58 billion in 2026 and USD 136.79 billion by 2035, a 34.52% growth rate (Precedence Research, Oct 2026).
• Mordor Intelligence estimated USD 6.90 billion in 2025 and a 42.83% growth rate to 2030 (Mordor Intelligence, Nov 2025).
• MarketsandMarkets projected a 17.5% growth rate to 2030 (MarketsandMarkets, Nov 2024).

These are research-firm estimates with different definitions of what counts as AI in education. Precedence says so itself: some estimates capture only dedicated AI education software and others include embedded AI revenue (Precedence Research, Oct 2026). Treat them as ranges and name the firm.

Is venture money still flowing into edtech?

Education venture capital fell 72% from $10.6 billion in 2022 to $3.0 billion in 2023, a further 22% to $2.3 billion in 2024, then rose 11% to $2.6 billion in 2025 (HolonIQ, Apr 2026).

• Cumulative education venture funding from 2010 to Q1 2026 was $29.7 billion in China and $29.2 billion in the US, against $11.8 billion in Europe and $10.0 billion in India (HolonIQ, Apr 2026).
• Workforce training took 38% of 2025 deal volume, K-12 36%, post-secondary 22% and early childhood 4% (HolonIQ, Feb 2026).
• Nearly 40% of 2025 venture deals were above $5 million, and Workday bought the AI learning platform Sana for $1.1 billion (HolonIQ, Feb 2026).
• Q1 2026 funding was $512 million across 63 deals, 24% lower in value and 10% lower in volume than Q1 2025 (HolonIQ, Apr 2026).
• Workforce training took more than 70% of Q1 2026 venture funding, and Preply's $150 million raise was the quarter's largest round (HolonIQ, Apr 2026).
• Education M&A hit 116 transactions in Q1 2026, led by the Coursera and Udemy merger valued at over $2.5 billion (HolonIQ, Apr 2026).

Capital is following job-linked and operational tools rather than classroom AI alone.

What has AI done to edtech companies?

Chegg, an early homework-help company, says in its annual report that Google's AI Overviews and students' use of free and paid generative AI have cut traffic to its site and new subscriptions (Chegg, Mar 2026).

• Google rolled out AI Overviews broadly in August 2024, and Chegg says their prevalence will only continue to increase (Chegg, Mar 2026).
• Its May 2025 and October 2025 restructurings cut about 640 employees, around 56% of its then-current workforce (Chegg, Mar 2026).
• It recorded $51.5 million of restructuring charges in 2025 and ended the year with 595 employees (Chegg, Mar 2026).

An answer shown in search results or a chatbot reply does not send the visit to the publisher.

Who is being left behind?

Access to AI is uneven by class, race, gender, place and language, and the gaps often point in different directions.

• Among UK undergraduates, 59% from social grades A, B and C1 used ChatGPT-style text tools, against 51% from C2, D and E, and 20% against 13% used AI to write code (HEPI, Mar 2026).
• Men were about 12 percentage points more likely than women to have used AI at school before university (HEPI, Mar 2026).
• Two-thirds of US high school students whose parents hold graduate degrees used generative AI frequently for schoolwork (College Board, Dec 2025).
• Black and Hispanic high school students were the most frequent users at 65%, against 58% of Asian students and 55% of White students (College Board, Dec 2025).
• About six in ten Black or Hispanic teens used chatbots for schoolwork help, against roughly half of White teens (Pew Research Center, Feb 2026).
• One in five teens in households earning under $30,000 did all or most of their schoolwork with chatbots, against 7% in higher-earning households (Pew Research Center, Feb 2026).
• 37% of Black teens felt extremely or very confident using chatbots, against 26% of Hispanic teens and 23% of White teens (Pew Research Center, Feb 2026).
• US teachers in suburban schools used AI at 65%, against 58% in urban and 57% in town and rural schools (Gallup, Jun 2025).
• 13% of Black teens said their teachers mostly allow generative AI, against 7% of White teens and 7% of Latino teens (Common Sense Media, Sep 2024).

The pattern is mixed. Wealthier and male students use more tools, and Black and Hispanic students lean on chatbots more for schoolwork, while the false-flag data above show the cost of detection errors can fall on the same groups.

Methodology and Sources

We included a number only after opening the page that publishes it and confirming the figure and its date. Figures found in roundups were traced to the original survey, paper, filing or release, or dropped. Survey figures describe different populations, so the text names each one: Pew Research Center, RAND, Gallup, the College Board and Common Sense Media survey US students and teachers, HEPI covers UK undergraduates, the OECD's TALIS covers lower secondary teachers in 55 systems, and the Digital Education Council surveys students and faculty in several regions. Vendor and company data from Anthropic, Turnitin and Chegg are labelled as such, and the two arXiv trials are marked as preprints. Market-size figures are research-firm estimates with differing definitions. The Digital Education Council's 2026 global survey page carries no publication date; we dated it June 2026 from its position in the council's publication list and its first web-archive capture on 1 July 2026. We refresh the page monthly.

Sources

• Anthropic, Anthropic Education Report: How university students use Claude (vendor data)
• arXiv, Henkel et al., Effective and Scalable Math Support: Evidence on the Impact of an AI-Tutor on Math Achievement in Ghana (preprint)
• arXiv, Wang et al., Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise (preprint)
• British Journal of Educational Technology, Fan et al., Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance
• Chegg, Form 10-K for fiscal year 2025, filed with the SEC
• College Board, Follow-up report on AI in high schools: GenAI use and policy vary by academic achievement, attainment, grade and context
• College Board, New research: faculty express near-universal concern that student AI use undermines learning
• College Board, New Research: Majority of High School Students Use Generative AI for Schoolwork
• Common Sense Media, The Dawn of the AI Era: Teens, Parents, and the Adoption of Generative AI at Home and School (PDF)
• Digital Education Council, AI in Higher Education Global Survey 2026
• Digital Education Council, What Faculty Want: Key Results from the Global AI Faculty Survey 2025
• Digital Education Council, What Students Want: Key Results from DEC Global AI Student Survey 2024
• EUR-Lex, Regulation (EU) 2024/1689 (the EU AI Act)
• Federal Register, Final Priority and Definitions: Secretary's Supplemental Priority on Advancing Artificial Intelligence in Education
• Gallup, Three in 10 Teachers Use AI Weekly, Saving Six Weeks a Year
• Grand View Research, AI In Education Market Size, Share And Growth Report, 2033 (research-firm estimate)
• HEPI, Student Generative AI Survey 2025
• HEPI, Student Generative AI Survey 2026, full report (PDF)
• HEPI, Student Generative Artificial Intelligence Survey 2026 (HEPI Report 199)
• HolonIQ, $512M in Q1 signals a slow start to 2026
• HolonIQ, EdTech hits $2.6B in investment as the market stabilizes
• Humanities and Social Sciences Communications, Wang and Fan, The effect of ChatGPT on students' learning performance, learning perception, and higher-order thinking: a meta-analysis
• International Journal for Educational Integrity, Weber-Wulff et al., Testing of detection tools for AI-generated text
• Journal of Academic Ethics, Perkins et al., Detection of GPT-4 generated text in higher education
• MarketsandMarkets, AI in Education Market (research-firm estimate)
• Mordor Intelligence, AI in Education Market (research-firm estimate)
• OECD, Digital Education Outlook 2026, full report (PDF)
• OECD, Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education
• OECD, Results from TALIS 2024: The State of Teaching
• OpenAI, New AI classifier for indicating AI-written text (withdrawn July 2023)
• Patterns, Liang et al., GPT detectors are biased against non-native English writers
• Pew Research Center, A quarter of U.S. teachers say AI tools do more harm than good in K-12 education
• Pew Research Center, Demographic differences in how teens use and view AI
• Pew Research Center, How Teens Use and View AI
• PLOS ONE, Scarfe et al., A real-world test of artificial intelligence infiltration of a university examinations system
• PNAS, Bastani et al., Generative AI without guardrails can harm learning: Evidence from high school mathematics
• Precedence Research, AI in Education Market (research-firm estimate)
• RAND, AI Use in Schools Is Quickly Increasing but Guidance Lags Behind, full report (PDF)
• RAND, AI Use in Schools Is Quickly Increasing but Guidance Lags Behind: Findings from the RAND Survey Panels
• Scientific Reports, Kestin et al., AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting
• Stanford Graduate School of Education, What do AI chatbots really mean for students and cheating?
• Stanford HAI, AI Index Report 2026, Chapter 7: Education
• The Guardian, Revealed: thousands of UK university students caught cheating using AI (freedom of information investigation)
• The White House, Executive Order 14277, Advancing Artificial Intelligence Education for American Youth
• Turnitin, Understanding the false positive rate for sentences of our AI writing detection capability (vendor data)
• Tyton Partners, Time for Class 2025: Empowering Educators, Engaging Students
• UK Department for Education, Generative AI: product safety standards
• UK Department for Education, School and college voice: December 2025
• UK Department for Education, School and college voice: November 2024
• UNESCO, UNESCO survey: Less than 10% of schools and universities have formal guidance on AI
• Vanderbilt University, Guidance on AI detection and why we are disabling Turnitin's AI detector
• Walton Family Foundation and Gallup, Teaching for Tomorrow: Unlocking Six Weeks a Year With AI (PDF)
• World Bank, From Chalkboards to Chatbots: Evaluating the Impact of Generative AI on Learning Outcomes in Nigeria

Glossary

• Generative AI (GenAI): AI that writes text, images or code on request, such as ChatGPT, Gemini or Claude.
• K-12: kindergarten through grade 12, the US school years before college.
• TALIS: the OECD Teaching and Learning International Survey, which asks teachers and principals about their work.
• RCT: randomised controlled trial, a study that assigns people to groups by chance.
• Standard deviation (SD) and Hedges' g: measures of effect size. 0.2 is usually called small, 0.5 medium and 0.8 large.
• Preprint: a paper posted publicly before peer review.
• Metacognitive laziness: handing off planning and self-checking to a tool, which can weaken learning.
• False positive: human work wrongly flagged as AI-written.
• Edtech: education technology companies and products.
• Venture capital (VC) and M&A: investment in young companies, and mergers and acquisitions of existing ones.
• CAGR: compound annual growth rate, the yearly growth rate implied by a forecast.
• SEND: special educational needs and disabilities, the English term for learning support needs.
• HEPI: the Higher Education Policy Institute, a UK think tank.
• RAND American Teacher Panel: a nationally representative panel of US public school teachers.