AI in education statistics: what the primary surveys actually say

Disclosure: NOXI publishes this page and is the presentation tool named in its slide-building sections. Every statistic on the page comes from the primary survey, report or paper linked beside it, each opened on 15 October 2026; nothing here was commissioned or funded by those organisations.

Most pages that rank for this query list 50 to 80 numbers with no population, no year and often no study: the same "student AI use" appears as 66%, 86% or 92% on different lists, because each figure measures a different group in a different year. This page does the opposite. Every number links to the survey it comes from, with the population, sample, fieldwork dates and question scope, and the popular claims are traced back to where they started. It is a dated edition, built to be refreshed when Pew, HEPI, Gallup and the OECD publish their next waves.

AI in education statistics at a glance (checked 15 October 2026)

The twelve numbers below are the ones worth citing about AI in schools and universities in 2026; each comes from a named survey with a known population, sample and year. Read the "what it measures" column before quoting any of them: most errors on this topic come from attaching a correct number to the wrong group.

How many university students use AI?

Almost all of them in the UK: 95% of full-time UK undergraduates said they had used AI in at least one way this academic year, and 94% said they had used generative AI to help with assessed work, in the HEPI/Kortext Student Generative AI Survey 2026. The same survey found 92% any use in 2025 and 66% in 2024.

The Higher Education Policy Institute (HEPI) runs this survey each year with Kortext as sponsor. For the 2026 edition, Savanta polled 1,054 full-time undergraduates in December 2025; responses were weighted by gender, institution type and year of study, and the margin of error is about 3% (HEPI Report 199, Rose Stephenson and Charlotte Armstrong). The headline question asks which uses of AI students had made "this academic year", for any purpose, including hobbies and jobs, so 95% is broader than study use alone. The assessed-work question defines generative AI as tools that generate text, images and other content in response to prompts, with ChatGPT, Google Gemini and Microsoft Copilot as examples.

Source: HEPI Report 199, Figures 1, 4 and 25 and the accompanying text. The 2024 survey was polled by UCAS and the 2025 and 2026 surveys by Savanta, and new answer options are added each year, so HEPI itself advises comparing the trend with caution. The most common assessed-work uses in 2026 were asking an AI assistant to explain concepts (61%), summarising a relevant article (49%), suggesting research ideas (40%) and structuring thoughts (39%). Fewer students said they generated text than in 2025 (56% against 64%), which the authors suggest may reflect a move to task-specific tools rather than less reliance.

The survey also shows how students feel about it. Some 65% say assessment has changed significantly in response to AI, 49% believe AI has improved their student experience, and 68% say AI skills are essential to thrive in today's world, while only 48% feel their teaching staff are helping them develop those skills. About 15% use AI for friendship, company, advice or to tackle loneliness, a new option in 2026. For students choosing tools, our guide to AI presentation makers for students covers the presentation side.

How many teens use AI for schoolwork?

About half of US teens: 54% of teens aged 13 to 17 said they had used AI chatbots to get help with schoolwork, and 57% to search for information, in a Pew Research Center survey fielded from 25 September to 9 October 2025. Only about one in ten said they do all or most of their schoolwork with chatbots' help.

Pew surveyed 1,458 US teens and their parents online; Ipsos recruited the teens through parents in its probability-based KnowledgePanel, and the margin of error is plus or minus 3.3 percentage points (Pew Research Center, How Teens Use and View AI, 24 February 2026). Overall, 64% of teens say they use AI chatbots, including about three in ten who do so daily; fewer parents than teens report it: about half of parents say their teen uses chatbots.

Source: Pew Research Center, 2026. About four in ten teens or more have used chatbots to research a topic or solve maths problems, and 35% to edit something they wrote. About a quarter of all teens say chatbots have been extremely or very helpful for schoolwork, another 25% somewhat helpful, and 3% of little or no help.

Why 26% (2024) and 54% (2025) cannot be compared

The two Pew numbers answer different questions, so they are not a trend. In the autumn 2024 survey of 1,391 teens, 26% said they used ChatGPT for their schoolwork, up from 13% in 2023 (Pew, 15 January 2025). That question named one product. The 2025 survey asked about AI chatbots in general and about getting "help with schoolwork", a broader category. The defensible sentence is "26% of US teens used ChatGPT for schoolwork in 2024; in 2025, 54% had used any AI chatbot for schoolwork help". The 2024 survey also asked what teens find acceptable: 54% said it is acceptable to use ChatGPT to research new topics, 29% to solve maths problems and 18% to write essays, while 42% said essay writing is not acceptable.

How many teachers use AI? US teachers in 2025

Six in ten US public school teachers (60%) used an AI tool for their work in the 2024–25 school year, and 32% used AI at least weekly, according to the Walton Family Foundation and Gallup. Teachers who used AI weekly estimated that it saved them 5.9 hours a week, which Gallup equates to about six weeks over a school year.

The survey ran on the web from 18 March to 11 April 2025 with 2,232 US teachers in public K-12 schools, recruited from the RAND American Teacher Panel, a nationally representative probability-based panel; the margin of error is plus or minus 2.5 percentage points (Gallup, Three in 10 Teachers Use AI Weekly, Saving Six Weeks a Year, 24 June 2025). The full report, Teaching for Tomorrow: Unlocking Six Weeks a Year With AI, gives the split: 40% not using AI at all, 28% using it infrequently and 32% at least weekly.

Two cautions apply. The time saved is self-reported by weekly users, not measured, and it describes the third of teachers who use AI weekly, not all teachers. Attitudes were also cooler a year earlier: in Pew's autumn 2023 survey of 2,531 public K-12 teachers, 25% said AI tools do more harm than good in K-12 education, 6% more good than harm, 32% an equal mix and 35% were not sure; high school teachers were the most negative, at 35% (Pew, 15 May 2024).

Teachers worldwide: OECD TALIS 2024

On the OECD average, 36% of lower secondary teachers reported having used artificial intelligence in their work in the TALIS 2024 survey, published on 7 October 2025. Use ranged from about 75% in Singapore and the United Arab Emirates to 14% in France and 17% in Japan.

TALIS, the OECD Teaching and Learning International Survey, is the largest international survey of teachers and principals: in 2024, 280,000 educators in 55 education systems took part (OECD, Results from TALIS 2024: The State of Teaching). Its definition of AI is wider than chatbots: "the capacity for computers to perform tasks traditionally thought to involve human intelligence", which also covers speech recognition, learning analytics and adaptive tools. Teachers who use AI most often use it to learn about and summarise a topic (68%) or to generate lesson plans or activities (64%); only 25% use it to review data on student participation or performance and 26% to assess or grade student work (TALIS 2024, chapter 1).

Source: OECD, Results from TALIS 2024 country notes, for example United Arab Emirates, Saudi Arabia and United States (one note per system; we read all 55 on 15 October 2026). "Non-users" are teachers who had not used AI in their teaching in the 12 months before the survey. The OECD's own comparison column matters: the United States, at 43%, is classed as similar to the 36% average, so the difference is not statistically significant in the OECD's analysis. Among the Gulf systems, non-users in Saudi Arabia (64%) were far more likely than the OECD average (37%) to say their school lacks the infrastructure to use AI, while in the UAE the figure was 15%.

Teachers' views are mixed. Around half agree that AI helps them create or improve lesson plans, from 18% in France to 91% in Viet Nam, with 87% in the UAE, and seven in ten believe AI could let students misrepresent others' work as their own. The OECD notes that TALIS relies on self-reports shaped by social and cultural context, so cross-country comparisons should be read with caution.

AI policy in schools: how many teachers and students have guidance?

Few: only 18% of US public K-12 teachers receive any formal guidance on AI from school administrators, 48% receive informal guidance only and 34% receive none, according to a Walton Family Foundation and Gallup survey of 2,069 teachers from 9 February to 2 March 2026. In 2025, only 19% of US teachers worked at a school with an AI policy.

The 2026 figures come from Gallup's article Most Teachers Receive No Formal Guidance on AI Use (26 May 2026), based on the study Teaching for Tomorrow: Closing the Expectations Gap; teachers were again drawn from the RAND American Teacher Panel, with a margin of error of plus or minus 2.5 points. Guidance was measured across ten work tasks. It is thinnest where AI touches students most directly: 69% of teachers say they have no guidance on using AI for one-on-one instruction or tutoring, and 58% none for grading and student feedback.

Universities show a similar gap between use and support. In HEPI's 2026 survey, 37% of UK undergraduates agreed that their institution encourages them to use AI (28% agreed in 2025); 38% said their institution provides AI tools, while 50% think it should. The global picture was thinner still in 2023: a UNESCO survey of more than 450 schools and universities found that fewer than 10% had institutional policies or formal guidance on generative AI (UNESCO, 7 September 2023). The same UNESCO guidance sets an age limit of 13 for the use of AI tools in the classroom and calls for teacher training.

Cheating, integrity and AI detection: what the data says

The data show widespread worry about AI cheating and weak tools for proving it: 59% of US teens think students at their school use AI to cheat at least somewhat often (Pew, 2025), 12% of UK undergraduates say they have pasted AI-generated text directly into assessed work (HEPI, 2026), and OpenAI withdrew its own AI-text classifier in 2023 because of its low accuracy.

Each of these numbers measures something different. Pew's 59% is a perception of other students; among teens who have used chatbots for schoolwork, 76% say classmates use them to cheat at least sometimes, while 14% of all teens say it rarely or never happens and 15% are unsure. HEPI's 12% is a self-report of one specific practice, up from 3% in 2024 and 8% in 2025; whether it breaks the rules depends on the assignment. Students also fear the process: being accused of cheating was the reason most often given for being less likely to use AI (42% in 2026, down from 53% in 2025), and HEPI notes free-text answers expressing anxiety about false accusations. Among teachers, seven in ten across OECD systems believe AI could enable students to misrepresent others' work as their own (TALIS 2024).

The second study's authors caution against using detectors in evaluative or educational settings because they can penalise non-native English writers. The practical reading for a school: a detector score can start a conversation, but on its own it is not evidence of misconduct.

The myth ledger: six AI in education claims and their sources

Six claims about AI in education are repeated far more often than they are checked. The ledger below gives each claim, where it comes from, what the original source measured and a verdict; every origin was opened on 15 October 2026.

One more pattern deserves a line of its own: lists that quote 66%, 86%, 92% and 95% as if they were competing estimates of the same thing. Two of those numbers (66% and 92%) are earlier editions of the same HEPI survey, for 2024 and 2025, and 95% is the 2026 edition. We could not trace 86% to a primary source we could open, so we do not use it.

Which AI in education statistic to cite, sentence by sentence

Choose the statistic by the sentence you want to write, then copy the wording in the third column; most misquotes happen when a correct number is attached to the wrong claim. The table covers the ten sentences people most often want to write about AI in education.

What the numbers mean for a school of 60 teachers

Applied to a school of 60 teachers, the OECD average suggests about 22 would already use AI in their work and, at the US rate, about 11 would have formal guidance. These are population rates used as an illustration, not a prediction for any specific school; a quick anonymous staff survey will always beat them.

Three practical takeaways follow from the data. First, guidance is the gap, not use: write down what is allowed for planning, feedback and grading, because those are the tasks with the least guidance today. Second, decide on integrity before you buy tools: detectors cannot settle a case alone, so design assessments that show process, such as drafts, oral checks and in-class work. Third, think about data before uploading anything: student work and names need the same care in an AI tool as anywhere else, and our guide to confidential material in AI presentation tools lists the questions to ask. For staff who want to start with lesson materials, our review of AI presentation tools for teachers compares the options.

How to turn AI in education statistics into sourced slides

A statistic on a slide needs four things besides the number: the population, the year, the exact measure and a source line. "54% of US teens aged 13–17 have used AI chatbots for schoolwork help (Pew, 2025)" is a citable statement; "most students use AI" is not, because it has no population, no year, no measure and no source.

For a staff training session, the eight-slide outline below turns this page into a deck. Each data slide carries one finding and its source line; our guide to AI presentations with research and citations covers source lines in depth, how to present data covers chart choice, and the accessibility review covers contrast and text size.

What we left out and why

We left out every figure we could not trace to a named survey or study with a population, sample and year, and every figure whose primary source we could not open on 15 October 2026. That rule removed most of the numbers that circulate on statistics lists.

• Market-size forecasts for "AI in education", which are commercial estimates with undisclosed methods; they describe vendor revenue, not what happens in classrooms.
• Learning-outcome claims such as "students using AI score 54% higher", for which we found no primary study.
• Vendor surveys without published methods, including headline figures behind download forms. The Digital Education Council's Global AI Student Survey 2024 (3,839 responses in 16 countries) is excluded for that reason: its public page does not show its headline figures.
• Detector vendors' own accuracy claims, which are marketing rather than independent evaluation.
• Single-country figures presented as global: where only a US or UK number exists, the page says so.

Key terms

How to cite this page

Cite the primary source for any single number, and cite this page when you use its compilation, the myth ledger, the citation table or the school calculations. Always include the "last checked" date, because HEPI, Pew and Gallup publish new waves every year.

Methodology, refresh calendar and update log

Every figure on this page was read from its primary source on 15 October 2026: HEPI Report 199, Pew Research Center's 2026 report and its 2024 and 2025 short reads, Gallup's 2025 and 2026 articles and the 2025 Teaching for Tomorrow report, the OECD's TALIS 2024 report chapter and all 55 country notes, UNESCO's 2023 announcement, OpenAI's classifier page and the peer-reviewed detector study by Liang and colleagues. The method is the same as on our public speaking statistics page.

• Percentages are reported as each source publishes them. Pew's "about 1 in 10" and "about three in ten" are Pew's own wording.
• HEPI's encouragement figure appears as 37% agree in the report's summary (11% strongly agree plus 26% agree in Figure 17) and as 36% in one passage of the text; we use the summary.
• School calculations are our arithmetic from published rates, shown in full; they are illustrations, not estimates for any real school.
• Never merged: Pew (US teens 13–17), HEPI (UK full-time undergraduates), Gallup (US public K-12 teachers) and TALIS (lower secondary teachers in 55 systems) measure different people, so the page never draws a trend line between them.

Update log. 15 October 2026: first edition, with HEPI 2026, Pew's February 2026 report, Gallup–Walton's May 2026 release and OECD TALIS 2024 as the latest published waves. When any of them publishes a new wave, the key numbers table, the figures and the quick answer will be updated on this URL and the change recorded here.

Three AI prompts for a sourced staff-training deck

The three prompts below give the model the numbers and the source lines, so it arranges evidence rather than inventing it; keep the brackets for your own school, audience and length.

How to build a sourced staff-training deck in NOXI

NOXI builds a full deck from a prompt or from your files and can match the style of a template your school already uses. It is free, with no card and no watermark; uploading files needs a free account, and a full deck uses about 25 of the 1,000 monthly credits. Your presentations, prompts and files are not used to train AI models.

Frequently asked questions

What percentage of students use AI?

It depends on the group. In the UK, 95% of full-time undergraduates used AI in at least one way in the 2025–26 academic year and 94% used generative AI for assessed work (HEPI/Kortext, December 2025, n = 1,054). In the US, 64% of teens aged 13–17 use AI chatbots and 54% have used them for schoolwork help (Pew, autumn 2025). There is no single global figure from a primary survey.

How many students use ChatGPT for school?

In Pew's autumn 2024 survey, 26% of US teens aged 13–17 said they used ChatGPT for schoolwork, up from 13% in 2023. Pew's 2025 survey asked about AI chatbots in general instead: 54% had used one to get help with schoolwork. The two figures come from different questions, so they should not be presented as a trend. HEPI found 56% of UK undergraduates generated text with tools such as ChatGPT in 2026.

How many teachers use AI?

On the OECD average, 36% of lower secondary teachers reported using AI in their work in TALIS 2024. In the US, 60% of public K-12 teachers used an AI tool for work in the 2024–25 school year and 32% used AI at least weekly, according to a Gallup and Walton Family Foundation survey of 2,232 teachers in spring 2025. Weekly users estimated saving 5.9 hours a week.

Is AI use in education increasing?

Yes, where the same survey repeats the same question. HEPI's survey of UK undergraduates found any AI use rising from 66% in 2024 to 92% in 2025 and 95% in 2026, and the share whose institution provides AI tools rising from 9% to 23% to 38%. Pew found ChatGPT use for schoolwork among US teens doubled from 13% in 2023 to 26% in 2024. Comparisons across different surveys are not trends.

Do schools have AI policies?

Mostly not formal ones. Only 18% of US public K-12 teachers received formal AI guidance from administrators in early 2026, 48% only informal guidance and 34% none (Gallup and Walton Family Foundation, n = 2,069). In 2025, 19% of US teachers worked at a school with an AI policy. A 2023 UNESCO survey of over 450 schools and universities found fewer than 10% had formal guidance.

Are AI detectors accurate?

Not accurate enough to be used alone. OpenAI withdrew its own AI-text classifier on 20 July 2023 because of its low accuracy: it identified 26% of AI-written text and wrongly flagged human text 9% of the time. In a 2023 study in Patterns, seven detectors misclassified human-written TOEFL essays by non-native English writers as AI-generated at an average rate of about 61%.

What do students use AI for?

Mostly to understand and organise material. Among UK undergraduates in 2026, 61% used an AI assistant to explain concepts for assessed work, 49% to summarise an article, 40% to suggest research ideas and 39% to structure their thoughts; 12% included AI text directly (HEPI). Among US teens, about four in ten or more used chatbots to research a topic or solve maths problems, and 35% to edit their writing (Pew).

Which countries' teachers use AI the most?

In OECD TALIS 2024, about 75% of lower secondary teachers in Singapore and the United Arab Emirates reported using AI in their work, followed by New Zealand (69%), Australia (66%) and Viet Nam (64%). The lowest shares were in France (14%) and Japan (17%). Bahrain reported 52%, Saudi Arabia and the United States 43%, against an OECD average of 36%.

Where do AI in education statistics come from?

The most reliable current figures come from four repeated surveys: HEPI/Kortext for UK undergraduates, Pew Research Center for US teens, Gallup and the Walton Family Foundation for US public school teachers, and the OECD's TALIS for lower secondary teachers in 55 systems. Each publishes its population, sample, fieldwork dates and questions, which is what makes a number citable.

Sources & methodology

All sources were opened and checked on 15 October 2026. The school calculations are our arithmetic from the published rates, shown above. Mariam and her school are fictional.

• [1] 95% any AI use, 94% for assessed work, 12% AI text included directly, 2024–2026 comparisons, 37%/36% encouragement, 38% provision, 68% and 48% skills, 42% fear of accusation; 1,054 full-time UK undergraduates, Savanta, December 2025 — Rose Stephenson and Charlotte Armstrong, Student Generative AI Survey 2026, HEPI Report 199, Higher Education Policy Institute and Kortext (March 2026); read via the Internet Archive copy of the official PDF (captured 2 October 2026).
• [2] 64% use chatbots, 57% search, 54% schoolwork help, 1 in 10 all or most schoolwork, 59% cheating perception, 76%, 14% and 15% splits; 1,458 US teens 13–17 and parents, 25 September–9 October 2025, margin of error ±3.3 points — Colleen McClain, Monica Anderson, Olivia Sidoti and William Bishop, How Teens Use and View AI, Pew Research Center (24 February 2026).
• [3] 26% of US teens used ChatGPT for schoolwork in 2024, 13% in 2023, acceptability of uses; 1,391 teens, 18 September–10 October 2024 — Pew Research Center, About a quarter of U.S. teens have used ChatGPT for schoolwork, double the share in 2023 (15 January 2025).
• [4] Teachers' views of AI in K-12 education, 2,531 public K-12 teachers, October–November 2023 — Pew Research Center, A quarter of U.S. teachers say AI tools do more harm than good in K-12 education (15 May 2024).
• [5] 60% used AI for work, 32% weekly, 5.9 hours saved by weekly users, six weeks a year; 2,232 US public K-12 teachers, 18 March–11 April 2025, ±2.5 points — Andrea Malek Ash, Three in 10 Teachers Use AI Weekly, Saving Six Weeks a Year, Gallup (24 June 2025).
• [6] 40%/28%/32% use split, 19% at schools with an AI policy, 2.3 vs 1.7 hours, 2.9 hours for monthly users, the 220.66-hour calculation, 40% favour and 28% oppose — Walton Family Foundation and Gallup, Teaching for Tomorrow: Unlocking Six Weeks a Year With AI (2025).
• [7] 18% formal, 48% informal only, 34% no guidance, 69% and 58% by task; 2,069 US public K-12 teachers, 9 February–2 March 2026, ±2.5 points — Andrea Malek Ash and Katherine Senseman, Most Teachers Receive No Formal Guidance on AI Use, Gallup (26 May 2026).
• [8] 280,000 educators in 55 education systems; Singapore and UAE around 75% — OECD, Results from TALIS 2024: The State of Teaching (7 October 2025).
• [9] OECD average AI use, uses of AI, AI definition, lesson-plan agreement, seven in ten on misrepresentation, self-report caution — OECD, Teaching for today's world, Results from TALIS 2024, chapter 1 (2025).
• [10] Country figures for AI use, lack of knowledge and skills and infrastructure, including the United Arab Emirates, Saudi Arabia, Bahrain and the United States — OECD, Results from TALIS 2024 country notes (7 October 2025), one per system, for example United Arab Emirates and Saudi Arabia.
• [11] Age limit of 13 for AI tools in the classroom; survey of over 450 schools and universities, fewer than 10% with policies or formal guidance — UNESCO, Governments must quickly regulate Generative AI in schools (7 September 2023); guidance page — UNESCO, Guidance for generative AI in education and research.
• [12] Classifier withdrawn on 20 July 2023 for low accuracy; 26% true positives and 9% false positives on a challenge set — OpenAI, New AI classifier for indicating AI-written text (31 January 2023, updated 20 July 2023).
• [13] Seven detectors, 91 TOEFL essays, 61.22% average false-positive rate, 18 and 89 essays flagged — Weixin Liang, Mert Yuksekgonul, Yining Mao, Eric Wu and James Zou, GPT detectors are biased against non-native English writers, Patterns 4(7), 100779 (July 2023); figures from the authors' arXiv version.
• [14] Digital Education Council survey size (3,839 responses, 16 countries), cited only to explain its exclusion — Digital Education Council, Global AI Student Survey 2024 (2 August 2024).

Turn these numbers into a staff session in NOXI, free: one statistic per slide, a title that states the finding and a source line under every figure, in the style of your school's template, exported as an editable PowerPoint or PDF.