Generative AI has changed how college students work and how colleges police academic integrity. The data below brings together the most widely cited figures on student AI use, AI detection, detector accuracy, and misconduct cases, with the source and year for each.
Key takeaways:
- Most college students now use generative AI for coursework, but far fewer report using it to write full assignments.
- AI detectors flag a meaningful share of student papers, and even their makers acknowledge false positives.
- Research has found that detectors can misclassify writing by non-native English speakers at high rates.
- Where cases are tracked, proven AI misconduct cases have risen sharply since 2023.
A note on the data: statistics in this area come from different surveys, countries, time periods, and definitions of “use,” and new data is published regularly. Each figure below is attributed to its source and year so readers can see where it comes from.
How Many College Students Use AI?
Short answer: Most do. In a 2025 Inside Higher Ed Student Voice survey, 85% of U.S. college students said they had used generative AI for coursework in the past year.
According to Inside Higher Ed’s 2025 survey, the most common uses were brainstorming ideas (55%), asking AI questions the way a student would ask a tutor (50%), and studying for exams or quizzes (46%). Fewer students reported using AI to complete assignments for them (25%) or to write full essays (19%). Community college students were less likely to report using AI: 21% of two-year students said they hadn’t used it in the past year, compared with 14% of four-year students.
Use has also grown quickly abroad. In the United Kingdom, the Higher Education Policy Institute’s 2025 Student Generative AI Survey found that 92% of full-time undergraduates used AI in some form, up from 66% in 2024, and that 88% had used it for assessments, up from 53%.
How Much Student Work Do AI Detectors Flag?
Short answer: In its first year, Turnitin’s AI detector reviewed over 200 million papers and found that about 11% contained at least 20% AI writing.
According to Turnitin’s April 2024 data, over 22 million papers (about 11%) had at least 20% AI writing present, and over six million (about 3%) had at least 80% AI writing present.
How Accurate Are AI Detectors?
Short answer: Not perfectly accurate, by their makers’ own account. Turnitin has reported a document false positive rate of less than 1% for documents with over 20% AI writing, and a higher rate of false positives below that threshold.
In a 2023 update, Turnitin’s chief product officer reported that documents with less than 20% detected AI writing had a higher incidence of false positives, which led the company to flag those scores with an asterisk. The same update described a sentence-level false positive rate of roughly 4%. Turnitin has also stated that its tool does not make a determination of misconduct, leaving that judgment to educators.
Concerns about false positives led several universities, including Vanderbilt, Michigan State, Northwestern, and the University of Texas at Austin, to decide not to use Turnitin’s AI detection feature. Vanderbilt’s reasoning was that even a false positive rate under 1% could mean hundreds of students wrongly flagged each year at a single school.
Are Some Students Flagged More Than Others?
Short answer: Research suggests yes. A 2023 Stanford study found that AI detectors misclassified essays by non-native English writers as AI-generated at an average rate of about 61%.
In the study by Liang and colleagues, published in the journal Patterns, seven widely used detectors were tested on 91 human-written TOEFL essays and 88 essays by U.S. eighth graders. The detectors were near-perfect on the U.S. essays but misclassified over half of the TOEFL essays as AI-generated, with an average false positive rate of 61.22%. All seven detectors agreed on 19.78% of the TOEFL essays, and 97.8% were flagged by at least one detector. The researchers linked the misclassifications to the lower “perplexity,” or predictability, of the essays’ language. Some detection companies have since reported improved results on similar tests, and detection technology continues to change.
How Many Students Get Caught for AI Cheating?
Short answer: Where data has been collected, proven cases have risen sharply. A Guardian investigation found almost 7,000 proven cases of AI cheating at UK universities in 2023-24, or 5.1 per 1,000 students, up from 1.6 per 1,000 the year before.
The investigation, based on Freedom of Information requests to UK universities, also found that figures through May of the following year pointed to roughly 7.5 proven cases per 1,000 students. Over the same period, proven cases of traditional plagiarism fell from 19 per 1,000 students to 15.2 in 2023-24. More than 27% of responding universities did not yet record AI misuse as its own category of misconduct, which suggests the reported numbers undercount the issue.
What the Numbers Mean for Students
Short answer: AI use is now common, detection tools are imperfect, and AI-related misconduct cases are increasing, which means more students, including students who did nothing wrong, are facing accusations.
Every one of those trends ends up in the same place: a student receiving an email from a professor or an academic integrity office. For more on that experience, see why AI use allegations are increasing even when students didn’t intend to cheat and why AI detection results alone can still trigger cheating cases.
Frequently Asked Questions
What percentage of college students use AI?
In a 2025 Inside Higher Ed Student Voice survey, 85% of U.S. college students said they had used generative AI for coursework in the past year.
How many college students use AI to write essays?
In the same 2025 survey, 19% of students said they had used AI to write full essays, and 25% said they had used it to complete assignments.
How often do AI detectors produce false positives?
Turnitin has reported a document false positive rate of less than 1% for documents with over 20% AI writing, a higher rate below that threshold, and a sentence-level rate of about 4%.
Are AI detectors biased against non-native English speakers?
A 2023 Stanford study found seven detectors misclassified essays by non-native English writers as AI-generated at an average rate of about 61%, while performing near-perfectly on essays by U.S. students.
How many students are caught cheating with AI?
A Guardian investigation found almost 7,000 proven AI cheating cases at UK universities in 2023-24, or 5.1 per 1,000 students, up from 1.6 per 1,000 the previous year.
Representation for Students Facing AI and Cheating Allegations
Statistics describe trends; every individual case turns on its own facts and on the policies of the institution involved. Attorney Richard Asselta is a student defense attorney who represents students facing academic misconduct allegations, including AI-related allegations, at colleges and universities nationwide. Call 855-338-5299.