How Many College Students Use AI

How Many College Students Use AI

In the HEPI 2026 survey, 95% of 1,054 full-time UK undergraduates said they used artificial intelligence (AI) in at least one way. In the guides we publish here at Chalkbox, we cite each survey with its year and population so you can quote it in a syllabus or committee memo. Tyton Partners' Time for Class 2026 found that 61% of US college students use AI at least weekly.

These figures differ because the surveys ask different questions of different groups. A question about any use, like HEPI's, produces a higher figure than a question about daily use, like Tyton's 32%. The populations differ too: HEPI surveyed UK undergraduates, while Tyton surveyed US college students.

Year-on-year HEPI data show rising use. The Higher Education Policy Institute (HEPI) found that UK undergraduate adoption rose from 66% in 2024 to 92% in the HEPI 2025 survey, before reaching 95% in 2026. Earlier US data from the Time for Class 2025 report by Tyton Partners showed 42% of students in introductory and developmental courses using generative AI weekly. Tyton's 2026 figure of 61% weekly use is not a like-for-like trend, because the 2025 report focused on those courses. Pew Research Center figures cover US teens aged 13 to 17, who are not college students. In Pew Research Center 2025 data, 26% of teens said they had used ChatGPT for schoolwork. In Pew Research Center 2026 findings, 54% said they had used chatbots "to get help with schoolwork."

Tasks Students Complete With AI Assistance

HEPI 2026 asked 1,054 full-time UK undergraduates which of these they had ever done when using generative AI to prepare assessed work:

• 61% used an AI assistant to explain concepts
• 49% used AI to summarise a relevant article
• 40% used an AI assistant to suggest ideas for research
• 39% used AI to structure their thoughts
• 36% used AI to search the internet
• 25% generated text with AI and edited it before submitting assessed work
• 12% included AI-generated text directly in a piece of assessed work
• 15% used AI "for friendship, company, advice or tackling loneliness" (a separate HEPI 2026 question, not about assessed work)

Including AI-generated text directly in assessed work is a minority practice, but it is growing. The HEPI 2026 report noted that the 12% figure represents an increase from 8% in 2025 and 3% in 2024. Meanwhile, 56% of UK undergraduates reported using AI to generate text in the 2025 to 2026 academic year, a decrease from 64% in the previous year. In HEPI 2026, 37% of UK undergraduates favour AI sources, 29% keep an equal balance between sources, and 33% lean towards traditional resources.

The Digital Education Council (DEC) Global AI Student Survey 2024 covered 3,839 students in 16 countries, and Campus Technology reported its figures. It found that 69% use AI to search for information. Checking grammar accounted for 42%, summarizing documents represented 33%, paraphrasing content reached 28%, and drafting initial text comprised 24%. That study found students use an average of 2.1 AI tools, with 66% naming ChatGPT, 25% using Grammarly, and 25% using Microsoft Copilot. Our guide to the pros and cons of AI in education covers where to draw the line between research support and unauthorized completion.

Faculty Use and Instructional Attitudes

In Tyton Partners' Time for Class 2026, 52% of US instructors reported using AI at least weekly, according to D2L's release. Tyton's 2025 report put weekly use at 30%, but among instructors in introductory and developmental courses, so the two are not a direct trend. Daily use stood at 25% among faculty, compared to 32% among students and 43% among campus administrators.

The Elon University Imagining the Digital Future Center and the American Association of Colleges and Universities surveyed 1,057 US college faculty (Elon/AAC&U faculty survey). The publisher calls it a non-scientific survey whose results are not generalizable to all college faculty. In it, 26% said they do not use generative AI tools at all, for personal or professional purposes. Furthermore, 32% specifically chose not to use generative AI for teaching and learning purposes. In that same survey, 87% of faculty stated they had created guidelines or policies for their students on AI use, and 69% reported addressing AI literacy in their instruction.

In the Elon/AAC&U survey, 76% of faculty reported believing "a lot" that generative AI tools may lead students to "develop an overreliance on Generative AI tools." Another 19% said "some." When asked about analytical rigor, 66% believed "a lot" that generative AI will diminish critical thinking skills, while 24% expressed "some" concern. Earlier research from Ithaka S+R in 2024 surveyed 2,654 instructors at US four-year institutions. It found that 72% had experimented with generative AI as an instructional tool, but only 14% felt confident using it in instruction. That study also showed 42% of instructors completely prohibited student use of generative AI, rising to 53% among humanities faculty.

The Digital Education Council Global AI Faculty Survey 2025 surveyed 1,681 faculty members across 52 institutions in 28 countries. It revealed that 83% of faculty are concerned about students' ability to evaluate AI outputs critically. In that survey, 86% see themselves using AI in teaching in the future. Still, 40% feel they are just beginning their AI literacy journey, and 54% believe current student evaluation methods are no longer adequate.

According to the Inside Higher Ed 2026 provosts survey, roughly seven in 10 provosts use AI in their own day-to-day work at least once per week. In the Inside Higher Ed 2025 provosts survey, 38% of provosts reported "significant faculty resistance to AI." In the same survey, 52% said "professors should not be forced to use or allow AI in their classrooms."

Institutional Policy and Strategic Planning

Only one in 10 US provosts reported a centralized AI strategy in Inside Higher Ed's 2026 survey. The table gives the other policy figures with their sources.

If your institution has no AI policy, an AI syllabus statement sets the rules for your course. In the Tyton Partners 2026 study, only 22% of faculty believed existing institution-wide policies were effective, compared to 44% of administrators. At institutions that ban AI entirely, 13% of faculty said their policy is effective. Where AI is integrated into assignments, 30% said so.

In the HEPI 2026 survey, 36% of UK undergraduates felt encouraged by their university to use AI, and 38% said their institution provides them with AI tools. Furthermore, while 68% of undergraduates viewed AI skills as essential to thrive, only 48% felt teaching staff helped them build those capabilities. In the Elon/AAC&U survey, 35% of faculty said their department has written specific guidelines on appropriate and inappropriate generative AI use, and 58% said it has not. For leaders drafting campus frameworks, our guide on AI policy for schools outlines structure for governance committees.

Academic Integrity and Detection Data

Proven cases of AI cheating at UK universities rose between 2022-23 and 2023-24. These figures are news reporting based on Freedom of Information requests. The Guardian 2025 misconduct investigation contacted 155 UK universities, and 131 provided some data. The investigation revealed almost 7,000 proven cases of AI cheating in the 2023 to 2024 academic year, or 5.1 cases per 1,000 students. That rate was up from 1.6 per 1,000 students in 2022 to 2023. The Guardian projected about 7.5 per 1,000 for 2024 to 2025, based on figures up to May. Confirmed traditional plagiarism dropped from 19 to 15.2 cases per 1,000 students between 2022-23 and 2023-24. More than 27% of responding universities did not yet record AI misuse as a separate misconduct category in 2023-24.

The non-scientific Elon/AAC&U survey of 1,057 US faculty asked about cheating and detection:

• 57% of faculty reported that cheating had increased "a lot" since generative AI arrived, with 21% reporting it had increased "a little"
• 33% had encountered academic integrity cases involving AI "a lot", and 40% encountered them "a few times"
• 31% used AI detection software on student assignments
• 34% noted their institution provided a paid detection tool subscription
• Only 2% rated current AI detection software "very effective", while 21% rated it "somewhat effective", 33% rated it "not very effective", and 22% called it "not at all effective"
• 38% expected generative AI to harm the value of academic degrees "some", with 36% expecting it to harm degree value "a lot"

Our guide to whether AI detectors are accurate covers what a detection score can and cannot show. The Guardian report noted that when University of Reading researchers tested their own institution's assessment systems, 94% of AI-generated submissions went undetected by existing university controls. For checks that do not rely on a detector score, see our overview of how do professors check for AI.

The HEPI 2026 report found that 42% of UK undergraduates said being accused of cheating by their institution would make them less likely to use AI, down from 53% in 2025. In the Tyton Partners 2026 study, faculty citing student cheating as a top challenge rose from 36% in 2024 to 55% in 2026. In the same study, 47% of US faculty had redesigned or were modifying assessments because of AI. That 47% splits into 24% redesigning around AI and 23% reverting to in-class or proctored formats.

Using Higher Education AI Statistics in Campus Governance

Faculty committees and curriculum leaders must handle survey percentages carefully when preparing course policies or administrative reports. A single percentage with no survey, country or population attached misleads a committee about student AI use.

When writing a departmental memo or syllabus policy, follow four citation standards:

1. Identify the surveyed student population explicitly: distinguish UK undergraduates (HEPI) from US survey groups (Tyton Partners) or global cohorts (Digital Education Council).
2. Name the frequency being measured: daily use (32%, Tyton 2026), weekly use (61%, Tyton 2026) and use in at least one way (95%, HEPI 2026) answer different questions.
3. State the publication date and survey year: reference whether data reflect recent 2026 fieldwork or earlier baseline studies from 2024 and 2025.
4. Avoid treating non-scientific samples as generalizable: note methodology limitations, such as the Elon/AAC&U faculty survey's non-scientific sample of 1,057 US college faculty.

If you assume that every student uses chatbots to bypass writing, you might default to blanket bans that prove difficult to enforce. If you assume students never use AI, you miss opportunities to teach critical evaluation. To adapt your coursework to these verified figures, review our guide on designing AI-resistant assignments to structure assessments that evaluate authentic student understanding.