AI cheating and plagiarism statistics: what the surveys and studies actually say

Key figures at a glance

Student use of generative AI is now the norm in the surveys below, while the share of students who hand in AI text as their own is much smaller. Recorded AI misconduct is rising, and independent studies keep finding that AI detectors make mistakes. Each row links to the original source, and the sample column shows who was asked.

How many students use AI

US teenagers (Pew Research Center)

Pew is the most-cited source on younger students because it repeats its questions over time. In its survey of 1,391 US teens aged 13 to 17, run from September 18 to October 10, 2024, 26% said they had used ChatGPT for schoolwork, up from 13% in 2023. In the same survey, 54% said it is acceptable to use ChatGPT to research new topics, 29% to solve math problems and 18% to write essays.

Pew’s February 2026 report, How Teens Use and View AI, asked about AI chatbots in general rather than ChatGPT alone. It surveyed 1,458 US teens and their parents from September 25 to October 9, 2025 (margin of error plus or minus 3.3 percentage points). 64% of teens said they use AI chatbots, and 54% said they had used them to get help with schoolwork. One in ten said they do all or most of their schoolwork with a chatbot’s help, 21% some, 23% a little, and 45% had not used chatbots this way.

UK undergraduates (HEPI and Kortext)

The Higher Education Policy Institute has run the same survey with Kortext three years in a row. Its Student Generative AI Survey 2026 was conducted by Savanta in December 2025 with 1,054 full-time UK undergraduates. 95% reported using AI in at least one way and 94% said they use generative AI to help with assessed work. The 2025 edition (1,041 students) reported that the share using generative AI tools for assessments had jumped from 53% in 2024 to 88% in 2025.

US college students and global surveys

• Inside Higher Ed and Generation Lab: in a July 2025 survey of 1,047 students at 166 two- and four-year US institutions (margin of error plus or minus 3 points), 85% said they had used generative AI for coursework in the last year. The most common uses were brainstorming ideas (55%) and asking it questions as a tutor would (50%). Fewer reported using it to complete assignments for them (25%) or to write full essays (19%).
• Tyton Partners: its Time for Class 2025 report, based on over 3,300 students, instructors and administrators at more than 900 US colleges, says weekly generative AI use “now spans 42% of students and 30% of instructors”. Note this is weekly use, a stricter measure than “ever used”.
• Chegg Global Student Survey 2025: Yonder Consulting interviewed 11,706 undergraduates aged 18 to 21 in 15 countries between October 1 and 23, 2024. 80% said they had used GenAI tools to support their university studies. Chegg sells study products, which is worth keeping in mind.

AI and cheating: what the data shows

Proven cases at UK universities (FOI data)

The clearest count of actual misconduct comes from a Guardian investigation published on June 15, 2025. The newspaper sent Freedom of Information requests to 155 UK universities, and 131 provided some data. It found almost 7,000 proven cases of cheating using AI tools in 2023-24, equal to 5.1 per 1,000 students, up from 1.6 per 1,000 in 2022-23. Over the same period, confirmed cases of traditional plagiarism fell from 19 to 15.2 per 1,000 students.

The Guardian also reported partial-year figures suggesting about 7.5 proven AI cases per 1,000 students in 2024-25. That is a projection from data up to May 2025, not a final count. More than 27% of responding universities did not yet record AI misuse as a separate category of misconduct in 2023-24, so these totals do not capture every case those universities dealt with.

Estimated AI cheating at US research universities (Science, 2026)

The largest estimate so far is a policy article published in Science on May 21, 2026 by Igor Chirikov (UC Berkeley), Ivan Smirnov (University of Technology Sydney) and René Kizilcec (Cornell). It uses survey data from 95,513 students in a representative sample of 20 major public research-intensive US universities, collected through the Student Experience in the Research University (SERU) undergraduate survey in the 2023-24 academic year. Because students may not admit cheating when asked directly, the researchers used a list randomization experiment: students said how many statements on a short list applied to them, but not which ones, and a statement about cheating was included on some versions of the survey only.

According to Cornell’s release on the study, about one-third of students regularly used generative AI when completing assignments, and 9% had used it to cheat. Daily users had the highest estimated rate of cheating, 26%, compared with 7% for monthly users. Overall, 37% reported using AI at least monthly. The article itself is behind a paywall; we read its abstract and the Cornell release, and the percentages here come from that release.

What students say about AI cheating

• Perceived cheating (US teens): in Pew’s fall 2025 survey, 59% of teens said students at their school use AI chatbots to cheat on their schoolwork at least somewhat often (34% extremely or very often, 25% somewhat often). 14% said rarely or never and 15% were not sure. Among teens who had used chatbots for schoolwork, 76% said cheating with chatbots happens at least sometimes. This measures what teens believe about their peers, not their own behavior.
• Pasting AI text into assessed work (UK): HEPI’s 2026 survey found the share of students who “included AI-generated text directly in a piece of assessed work” rose to 12%, from 8% in 2025 and 3% in 2024 (the 2026 report gives all three years). This is a separate answer option from using AI to generate text that students then edited: HEPI’s 2025 report put “AI-generated and edited text” at 18%, a figure its 2025 summary also describes as text included directly, so you may see 18% quoted for 2025. HEPI notes many students are now encouraged or required to use AI in assessments, so not all of this is misconduct.
• Why students say peers break AI rules (US): in the Inside Higher Ed and Generation Lab survey, pressure to get good grades was the top reason given (37%), and 97% said institutions should respond to academic integrity threats in the age of generative AI.

How much AI writing Turnitin finds

Turnitin published its own counts in an April 9, 2024 press release. Of over 200 million papers reviewed since its detector launched in April 2023 (as of March 21, 2024), over 22 million, about 11%, had at least 20% AI writing present, and over six million, about 3%, had at least 80%. These are Turnitin’s detector results, not confirmed cases. In an earlier update, Turnitin said its statistics include assignments where educators may have allowed or assigned AI use, and that it does not separate those out.

Can AI work slip through unnoticed?

In a blind study published in PLOS ONE in June 2024, Scarfe and colleagues submitted 100% AI-written answers into the exam system of five undergraduate psychology modules at a UK university (the Guardian names it as the University of Reading). 94% of the AI submissions were not detected, and on average they received grades half a grade boundary higher than those of real students.

What teachers and institutions report

The Center for Democracy & Technology surveys US K-12 teachers each year. Its March 2024 release, covering the 2023-24 school year and based on nationally representative surveys of K-12 public school teachers, reported that:

• 68% of teachers report using an AI content detection tool regularly;
• 64% say students at their school have gotten in trouble or faced negative consequences for using, or being accused of using, generative AI on a school assignment, a 16 percentage-point increase from the year before;
• 52% agree that generative AI has made them more distrustful of whether their students’ work is their own;
• only 28% say they have received guidance on how to respond if they suspect a student has used generative AI in ways that are not allowed.

CDT’s October 2025 release (806 teachers of grades 6 to 12, 1,030 students in grades 9 to 12 and 1,018 parents) found 85% of teachers and 86% of students had used AI in the 2024-25 school year, and seven in ten teachers worry that AI weakens important skills students need to learn.

On the university side, HEPI’s 2026 survey found only 36% of UK undergraduates feel their institution encourages them to use AI and 38% say they are given AI tools. For how teachers actually investigate a suspected case, see how teachers check for AI and do colleges use AI detectors.

Plagiarism statistics beyond AI

The longest-running source on cheating is research begun in 1990 by Donald McCabe, a founder of the International Center for Academic Integrity. These figures are older and self-reported, so read them as a baseline, not as current rates. According to ICAI’s facts and statistics page:

• McCabe’s original research and follow-up studies show that more than 60 percent of university students admit to cheating in some form.
• McCabe’s surveys of over 70,000 US high school students found 64 percent admitted to cheating on a test, 58 percent admitted to plagiarism, and 95 percent said they had taken part in some form of cheating.

In March 2020, ICAI researchers piloted an updated McCabe survey with 840 students at five institutions. ICAI describes this work as in preparation (Rettinger et al.), so treat it as preliminary.

For recorded cases rather than self-reports, the Guardian’s UK FOI data above is the most recent large dataset: confirmed traditional plagiarism fell from 19 to 15.2 per 1,000 students in 2023-24 as AI cases rose. If you are unsure where the line sits, our guides on what plagiarism is and whether using AI counts as plagiarism explain the definitions.

AI detector accuracy and false positives

Because AI misconduct is often first flagged by software, how accurate that software is matters. Here are the independent studies, with sample sizes, followed by one vendor’s own claim.

Vendor claim, clearly labelled: Turnitin’s AI detection FAQ says it aims to keep its false positive rate under 1% for documents with over 20% AI writing, which it describes as possibly flagging one in every 100 fully human-written documents. It says it tests each update on over 700,000 papers written before ChatGPT. This is the company’s own measurement and has not been independently confirmed by the studies above, which tested different tools at different dates.

Detectors change often, so a 2023 result may not describe today’s version of a tool. Our guides on AI detector false positives and whether AI detectors work go further, and how to prove you didn’t use AI covers what to do if your own work is flagged.

How to read these numbers

• “Use” is not “cheating”. Most surveys ask whether students used AI at all, for any purpose. Brainstorming, explaining a concept and pasting an essay all count as use. Many courses now allow or require some AI use.
• Definitions differ. “Ever used”, “used in the past year” and “use weekly” give very different numbers for the same students. Compare like with like.
• Self-reports undercount. Students are less likely to admit misconduct in a survey, and a question about peers (as in Pew’s) measures perception, not behavior.
• Recorded cases undercount too. FOI figures only include cases that were caught, proven and recorded as AI misuse, and more than a quarter of responding UK universities did not record that category in 2023-24.
• Detector counts are not verdicts. Turnitin’s share of papers with AI writing is a model’s estimate, including cases where AI use was allowed.
• Populations differ. US teens, UK undergraduates and a 15-country undergraduate panel cannot be combined into one global figure.
• The numbers move fast. HEPI’s assessment figure went from 53% to 88% in a year. Check the date before you quote.

Methodology, update log and citing this page

We list only figures we could read in the primary source, meaning the survey publisher’s own report or page, the journal article, or the newspaper’s own FOI article, on the date shown. Each figure is worded as the source frames it, without rounding that changes its meaning, and figures from different surveys are never combined. We left out figures we could only find in secondary coverage or behind a registration wall, such as the findings of the Digital Education Council’s 2024 global student survey. This page contains no data from Plagiarism Checker Plus.

Update log: first published October 17, 2026; all sources last checked October 17, 2026.