AI in education statistics 2026: 95% of UK students use AI, but only 38% get tools from their university

Students are ahead of their schools on AI in 2026

Each of these AI in education statistics names who was asked, where and when. A UK university survey and a US teen survey don't measure the same people. The date is when the survey ran, which comes before the day it was published.

Don't read these rows as one trend. HEPI polled UK undergraduates who study full time, Pew polled US teens and Gallup polled US teachers in public schools. Each asked its own question, too. "Use AI", "used chatbots for schoolwork" and "used an AI tool for work" aren't the same thing.

UK universities are catching up, but they're still 56 points behind their students. In HEPI's survey of 1,054 full-time undergraduates, run by Savanta in December 2025, 94% said they use generative AI for assessed work. Only 38% said their university provides generative AI tools. By our arithmetic from the three waves in HEPI's 2026 report, that gap was 44 points in 2024 and 66 in 2025. This is the first year it's shrunk.

Provision is what moved. In the last year, the share of students who get AI tools from their university rose 15 points, from 23% to 38%, while use for assessed work rose 5, from 89% to 94%. Two years ago it was 9%. Half of students (50%) think their university should provide those tools.

How many UK students use AI, and what for?

Almost all UK undergraduates now use AI. "AI use is now almost universal," HEPI wrote in its March 2026 report, and its three surveys with Kortext show how fast that happened.

UCAS polled 1,250 students in November 2023 for the first survey. Savanta polled 1,041 in December 2024 and 1,054 in December 2025, and HEPI weights each wave to be comparable. Every figure in this table comes from HEPI's 2026 report, which restates three of the 2025 numbers. HEPI's 2025 report gave 88% for assessed work (now 89%), 26% for university tools (24% in its own chart, now 23%) and 59% for changed assessment (now 56%). On the figures first published, the 2025 gap was 62 points. HEPI's 2025 summary said 18% had put AI text directly into their work; that was the share who generated text and edited it with a digital tool. The direct paste share was 8%.

The number that should worry universities is 12%. That's the share who say they've put text written by AI straight into a piece of assessed work. It was 3% in 2024. Fewer students now generate text with tools like ChatGPT (56%, down from 64%, though still up from 30% in 2024), and HEPI thinks they may be moving to more specific tools.

Students want help with it. 68% say AI skills are essential for their futures, but only 48% feel their teaching staff are helping them build those skills. 15% use AI for friendship, company or advice. And 23% say the environmental impact of AI makes them less likely to use it, a worry our AI data center energy consumption statistics put figures on.

How many US students use AI for school?

More than half of US teens and most US high school and college students use AI for school, though the share depends on who's asked and how. Pew Research Center found that 54% of US teens aged 13 to 17 have used chatbots to get help with schoolwork. It surveyed 1,458 teens from 25 September to 9 October 2025. One in ten said they do all or most of their schoolwork with a chatbot's help.

Pew's earlier question asked about ChatGPT alone. It found that 26% of teens used ChatGPT for schoolwork in 2024, double the 13% of 2023 (Pew, January 2025). The 2025 survey asked about chatbots in general, so 26% to 54% isn't one trend. In the same survey, 64% of US teens said they use chatbots, about three in ten daily, and 59% use ChatGPT (Pew, December 2025). Our AI chatbot market share page has the figures for adults.

College Board's surveys of high school students put use higher still. It found that 84% of US high school students used generative AI for schoolwork in May 2025, up from 79% in January. And 69% used ChatGPT for assignments and homework.

In college, AI is a weekly habit. The Lumina Foundation and Gallup found that 57% of US college students use AI in their coursework at least weekly, and about one in five use it daily. They surveyed 3,801 students from 2 to 31 October 2025. About half said their school discourages or prohibits AI. An Inside Higher Ed survey with Generation Lab, run in July 2025, found that 85% had used generative AI for coursework in the past year. Only 19% had used it to write a full essay.

The companies' own data points the same way. OpenAI said in February 2025 that more than a third of US 18 to 24 year olds use ChatGPT, and that about a quarter of their messages are about learning and school work. In April 2025, Anthropic said it had looked at 574,740 student conversations with Claude and found that about 47% were Direct, meaning the student wanted an answer with little back and forth. Computer science students, 5.4% of US degrees, made up more than a third of them.

Across 16 countries, the Digital Education Council surveyed 3,839 university students in 2024. It found that 86% use AI in their studies and 54% use it every week.

How many teachers use AI?

About six in ten US teachers and eight in ten UK teachers use AI for work, in the latest survey of each. Gallup and the Walton Family Foundation found that 60% of US public school teachers used an AI tool for work in the 2024 to 2025 school year, and 32% used one at least weekly. That survey covered 2,232 teachers in March and April 2025. Weekly users said AI saves them 5.9 hours a week. Gallup says that adds up to about six weeks a school year.

EdWeek Research Center's surveys show when the jump came. In 2023, 34% of US teachers said they used AI in their work at least a little. That slipped to 32% in 2024, then rose to 61% in 2025 (EdWeek, January 2026). RAND found that 53% of US English, math and science teachers used AI for planning or teaching in spring 2025.

In the UK, 80% of teachers use AI in their work, according to Bett UK and Lenovo's 2026 report. YouGov polled 1,033 teachers for it from 16 to 29 June 2026. Weekly use doubled in a year, from 26% to 52%, Bett said. Of the teachers who use it, 76% use it to make lesson plans, worksheets and other resources.

University teachers who use Claude mostly use it to build courses. Anthropic looked at about 74,000 educator conversations with Claude from May and June 2025, and 57% were about developing curricula. Grading was rarer. But when faculty did use it to grade, 48.9% of those chats handed the task to the AI outright, which Anthropic called concerning.

Do students use AI more than teachers?

Not clearly. We found three surveys that polled both groups, and only RAND's splits them by grade. At the same grade, US teachers were ahead. In spring 2025, 69% of high school English, math and science teachers used AI for planning or teaching. In winter 2025, 61% of high school students used it for schoolwork. In middle school it was 64% against 41%.

RAND's headline figures, 54% of students and 53% of teachers, look level only because the teacher figure includes elementary schools, where 42% used AI, and the student figure doesn't. Bigger gaps come from setting one survey against another. Put HEPI's 95% of UK undergraduates next to Bett's 80% of UK teachers and you'd get 15 points. But one covers universities and the other covers schools, so that gap doesn't mean anything.

Training is where the groups really differ. In Microsoft's survey of 3,345 people, 77% of students had no formal AI training, against 53% of educators. In the Latin American survey, 50% of students wanted AI feedback on their assignments, but only 19% of faculty gave it.

Do schools have AI policies and training?

Most schools don't have formal AI rules yet, though the share is growing. Gallup found that just 18% of US public school teachers have formal guidance on how to use AI, in a survey of 2,069 teachers from 9 February to 2 March 2026. About a third (34%) get no guidance at all. For AI tutoring, 69% have none.

In spring 2025, 45% of US principals told RAND their school or district had AI policies or guidance. In fall 2024, only 34% of teachers said theirs had a policy on AI and academic integrity. Only 35% of district leaders said they train students to use AI. More than 80% of students said their teachers hadn't explicitly taught them how to use it for schoolwork. College Board found that about two in five US schools or districts don't let students use generative AI, while about 13% encourage it in every class.

US teachers are getting more training. EdWeek Research Center found that 50% of US teachers had at least one training session on AI by fall 2025, up from 13% in 2023. The share with no training at all fell from almost 60% in October 2024 to 42% in the winter of 2025 to 2026 (EdWeek, May 2026). In the UK, 41% of teachers say their school hasn't trained them on AI, and only 33% say it has a policy on staff use.

Worldwide, UNESCO found in 2023 that fewer than 10% of more than 450 schools and universities had formal guidance on generative AI. By September 2025, 19% of respondents to its next survey said their university had a formal AI policy, and 42% said one was being written (UNESCO). That survey asked 400 people at UNESCO Chairs and networks in 90 countries, so it isn't a random sample.

The US government has made AI teaching a goal. President Trump signed Executive Order 14277, Advancing Artificial Intelligence Education for American Youth, on 23 April 2025. The Department of Education then made AI in education a priority it can apply to its discretionary grants. That priority was published as final on 13 April 2026 and took effect on 13 May.

Is AI cheating increasing?

Yes. Every count that's been repeated has gone up, though none of them measures cheating directly. In England's GCSE, AS and A level exams, AI misuse made up 75.0% of student plagiarism cases in summer 2025, up from 55.4% in 2024, Ofqual reported. That's 100 cases against 85, and 2.0% of all student malpractice.

At UK universities, a Guardian investigation found almost 7,000 proven cases of cheating with AI tools in 2023 to 2024. That's 5.1 per 1,000 students, up from 1.6 per 1,000 the year before. The Guardian got figures from 131 of the 155 universities it asked under the Freedom of Information Act. Early data pointed to about 7.5 per 1,000 for 2024 to 2025.

Turnitin's AI detector counts something else: how much of a submission looks AI written. From October 2025 to April 2026, 19.4% of US college submissions scored above 80% on its AI writing check, against 9.8% in the UK and 10.2% in Australia. In secondary schools it was 5% to 6%. In its first year, to March 2024, it reviewed more than 200 million papers and found about 3% with at least 80% AI writing (Turnitin). Those two counts don't cover the same students. A high score isn't proof of cheating, either, and many courses now allow AI.

Students and teachers think it's common. 59% of US teens told Pew that students at their school use AI to cheat at least somewhat often. In the UK, 30% of teachers had suspected a student of handing in unauthorized AI work in the past four weeks, and 57% in secondary schools had (Bett, June 2026). Half of US students told RAND they worry about being falsely accused of it.

Does AI help students learn?

It depends on how the tool is built, two experiments published in 2025 found. In a Harvard physics course in fall 2023, 194 students learned more than twice as much from an AI tutor built on the course's own teaching methods as from an active learning class. Most of them also took less time: a median of 49 minutes, against 60 minutes in class (Scientific Reports, 3 June 2025).

Plain chatbot access can do the opposite. In a field experiment with nearly a thousand math students at a high school in Turkey in fall 2023, those given a standard GPT-4 chat window scored 48% higher on practice problems. Once it was taken away, they scored 17% lower than students who never had it (PNAS, June 2025). A version with guardrails lifted practice scores by 127% and largely avoided the drop.

Parents worry about this more than school leaders do. RAND found in 2025 that 61% of US parents and 55% of high school students agreed that more AI use will harm students' critical thinking, against only 22% of district leaders.

Which AI companies sell to schools, and how many users do they claim?

OpenAI, Google and Anthropic all sell an education version of their chatbot. Each reports its reach in different units, so these figures aren't comparable. They're the companies' own claims, as of October 2026.

OpenAI called the Cal State deal the largest rollout of ChatGPT by any single organization when it was announced. None of the three publishes a total count of education users. For the user numbers behind each company, see our ChatGPT statistics, Claude statistics and Gemini statistics.

In the UK, the tool students most often say their university provides is ChatGPT, followed by Microsoft Copilot, HEPI found. Google sells Gemini for Education on its data protections, and our AI privacy statistics have the wider numbers on chatbot data. A student discount is a different thing from a campus deal. Our ChatGPT Plus student discount page has OpenAI's offers for individual students.

How we compiled these numbers

We opened every figure at its primary source on 7 October 2026. For a survey, that's the publisher's own report or release: HEPI, Pew Research Center, RAND, Gallup, College Board, EdWeek Research Center, Bett, the Digital Education Council, UNESCO and Ofqual. For a company, it's the company's own post: OpenAI, Google, Anthropic, Microsoft and Turnitin. US policy comes from the White House and the Federal Register. The two learning trials come from the journals that published them, Scientific Reports and PNAS. The university cheating figures are the Guardian's, and we name them as such. Every number, the wording at its source and the URL are in our research log.

We compared figures only inside one survey, or across one organization's surveys of students and teachers in the same school year, and the table names the question each group answered. The gaps and multiples on this page are our arithmetic from those figures. We left out every AI in education market size estimate. They come from paid market research reports that don't publish a method you can check. We also dropped any number we couldn't find at its origin, including a Turnitin share that appears in news coverage but not in Turnitin's own releases.

How to cite these AI in education statistics

You're free to use any figure or table on this page, as long as you link back to it. Please credit the original survey for the numbers that came from someone else, too. A plain credit line like this one works:

Source: Whizi Research, "AI in education statistics 2026: 95% of UK students use AI, but only 38% get tools from their university", whizi.io/resources/ai-in-education-statistics, updated October 2026.

More numbers from the same series: AI hallucination statistics, AI subscription statistics and LLM API pricing statistics.