Quick answer: How many students use ChatGPT for essays? Almost all university students now use generative AI somewhere in their coursework, but far fewer hand in text that a chatbot wrote.

In the largest UK tracker, the HEPI and Kortext Student Generative AI Survey 2026, 94% of full-time undergraduates said they use generative AI to help with assessed work, 56% said they use it to generate text with tools such as ChatGPT, and 12% admitted putting AI-generated text directly into a piece of assessed work. Among US teenagers, Pew Research Center found that 26% had used ChatGPT for schoolwork in 2024, double the 13% of 2023, while only 18% thought it was acceptable to use it to write essays.

Those numbers answer different questions, which is why headlines about how many students use ChatGPT for essays often seem to contradict each other. A student who asks a chatbot to explain a concept, another who pastes in a draft for grammar suggestions and a third who submits an essay the tool wrote are all “using ChatGPT for essays” in a loose sense, yet only the last one is what most integrity policies mean by misconduct. This page sorts the most reliable surveys by what they actually measured, explains the trend lines, and shows what the data mean for anyone writing a thesis or dissertation in 2026.

The headline numbers at a glance

The pattern is consistent across countries and methods: general AI use is close to universal, but wholesale submission of machine-written text is a minority behaviour. The gap between those two figures is where most of the interesting questions sit.

What the UK data show: near-universal use, minority substitution

The HEPI and Kortext survey is the most useful single source because it asks the same questions every year and separates types of use. Its 2026 report, based on 1,054 full-time undergraduates weighted by gender, institution type and year of study, puts the margin of error at about 3%.

Three figures matter most for essays:

• Using AI for assessed work at all: just over half of students in 2024, 89% in 2025 and 94% in 2026. The report concludes that almost all students now use AI to help complete their assessments, often because institutions encourage or require it.
• Generating text with tools such as ChatGPT: 64% in 2025, falling to 56% in 2026. The authors say the reason for the drop is unclear and suggest students may be moving to more task-specific tools. Generating text is still the single most common use.
• Putting AI-generated text directly into assessed work: 3% in 2024, 8% in 2025 and 12% in 2026. This is the closest the survey comes to measuring “ChatGPT wrote my essay”, and it has quadrupled in two years.

A further quarter of students in both 2025 and 2026 said they generated text with AI and then edited it before including it in assessed work. Whether that is allowed depends entirely on the assignment and the institution, which is the core of the policy problem described later on this page.

How students actually use AI on assessed work

When HEPI asked what students used AI for in assessed work in 2026, the most common answers were support tasks rather than writing:

• 61% used an AI assistant to explain concepts.
• 49% used AI to summarise a relevant article.
• 40% used it to suggest ideas for research.
• 39% used it to structure their thoughts.

The Digital Education Council’s global survey, as summarised by Campus Technology, shows a similar shape internationally: 69% of students used AI to search for information, 42% to check grammar, 33% to summarise documents, 28% to paraphrase a document and 24% to create a first draft. In the same survey 24% said they use AI daily, and the average student used 2.1 AI tools for their courses.

Read together, these surveys suggest that roughly one student in four uses AI to produce first-draft text at some point, while a much larger majority uses it as a tutor, search engine or editor. For thesis writers the distinction matters, because examiners and integrity panels judge what the tool did, not whether it was opened.

What the US data show: fast growth from a lower base

US data on college students is more fragmented, but the Pew Research Center teen survey gives a clean year-on-year comparison for the students who are now entering university. In its fall 2024 survey of 1,391 teens aged 13 to 17, 26% said they had used ChatGPT for schoolwork, twice the 13% recorded in 2023, while 73% had not used it that way.

Pew also asked which uses teens consider acceptable, and the answers draw the same line that university policies draw:

• 54% said it is acceptable to use ChatGPT to research new topics, and 9% said it is not.
• 29% said it is acceptable for solving math problems, and 28% said it is not.
• Only 18% said it is acceptable for writing essays, and 42% said it is not.

Between 15% and 21% were unsure about each task. In other words, even the generation that grew up with chatbots largely sees essay writing as the use that crosses a line, which helps explain why the share who admit submitting AI text stays far below the share who use AI at all.

What detection data add, and what they cannot tell you

Turnitin’s figures are often quoted as evidence of how many essays are written by ChatGPT. On the first anniversary of its AI writing detection feature, the company reported in an April 2024 press release that of more than 200 million papers reviewed up to 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%.

Those numbers describe what a classifier flagged, not what students confessed or what panels proved, so treat them as a different kind of evidence from the surveys above. A paper with a high score may have been drafted with AI, or it may be a human text the model misjudged; a paper with a low score may have been machine-written and then edited. The site’s breakdown of AI detection accuracy statistics covers the false-positive research in detail, and it is worth reading before you rely on any detector percentage as a prevalence estimate.

How many students use ChatGPT for essays: why the numbers differ

If you are citing these statistics in your own work, the most common error is comparing figures that measure different things. Four differences explain most of the apparent contradictions:

1. Population. HEPI surveys UK undergraduates, Pew surveys US teenagers, and the Digital Education Council mixes bachelor’s, master’s and doctoral students across 16 countries. Graduate researchers writing theses are a small slice of each sample.
2. Definition of use. “Used AI in at least one way” (95% in HEPI 2026) is a very different question from “included AI-generated text directly in assessed work” (12%).
3. Time. Adoption moved fast between 2023 and 2026, so a 2023 figure and a 2026 figure are not measuring the same world.
4. Self-report. Surveys rely on students admitting behaviour that may be prohibited, so the direct-inclusion figures are best read as a floor rather than a ceiling.

So when someone asks how many students use ChatGPT for essays, the honest answer depends on the question asked. When you write about prevalence in a literature review, state the survey, the population, the fieldwork date and the exact question wording for every figure, and avoid merging them into a single “most students cheat with ChatGPT” claim that none of the sources supports.

What students say worries them

The HEPI 2026 report notes that 65% of students say assessment has changed significantly in response to AI, and that some students describe anxiety about being falsely accused of misconduct. Students are also split on whether their institution encourages AI: 37% agree and 36% disagree, with agreement highest at Russell Group universities (39%, up from 26% the year before). And while 68% believe AI skills are essential to thrive in today’s world, only 48% feel their teaching staff are helping them develop those skills.

That combination, near-universal use with mixed signals from institutions, is why so many students cannot say with confidence whether what they are doing is allowed. The practical answer is to read your own programme’s rules rather than infer them from national statistics. The guide to whether ChatGPT is allowed for thesis writing walks through the policy categories most universities now use.

What this means if you are writing a thesis

Undergraduate essays are where most of the survey data come from, but the consequences are sharper at thesis level, where the document is longer, examined more closely and often defended orally. Four practical points follow from the evidence:

• Assume your examiners know the numbers. Supervisors read the same surveys and are alert to the minority who submit AI text. A thesis with no drafts, no notes and no version history is harder to defend if a question is raised.
• Declare what you used. Many universities now ask for an AI-use statement. The guide to writing an AI-use declaration for your thesis shows a format that covers tools, purposes and the parts you wrote yourself.
• Keep support uses separate from writing. Explaining a statistical test or summarising a paper you then read in full is the kind of use the surveys show is now mainstream. Generating paragraphs for submission is the use most policies restrict.
• Cite AI properly when you quote it. If AI output appears in your text as data or an example, cite it in the style your department uses; the page on how to cite AI-generated content covers APA, Harvard and MLA formats.

How to cite these statistics in your own work

If you are writing about AI in education, cite the original reports rather than news summaries wherever you can. Record the fieldwork date as well as the publication date, because the HEPI 2026 survey was fielded in December 2025 and the Pew teen survey in fall 2024. Quote the exact question when the wording drives the number, for example “included AI-generated text directly in a piece of assessed work” rather than “cheated with AI”. And say plainly when a figure comes from a vendor, as Turnitin’s detection data do, so that readers can weigh it accordingly.

For related guides, see academic misconduct statistics in higher education and how AI writing detectors work.

Where Tesify fits

The surveys show students want help with structure and understanding rather than a machine that writes for them. Start structuring your thesis in Tesify. It is built for that side of the work: it helps you structure and organize your thesis chapter by chapter, and 9,000+ students have written 15,000+ chapters with Tesify. Every word stays 100% written by you.

Frequently asked questions (FAQ)

What percentage of college students use ChatGPT?

In the HEPI and Kortext Student Generative AI Survey 2026, 56% of UK full-time undergraduates said they use AI to generate text with tools such as ChatGPT, and 95% use some form of AI. The Digital Education Council’s 2024 global survey reported ChatGPT as the most used tool, at 66% of students.

How many students submit essays written by AI?

The best available self-report figure is from HEPI 2026, where 12% of UK undergraduates said they had included AI-generated text directly in a piece of assessed work, up from 8% in 2025 and 3% in 2024. Because the behaviour is often prohibited, treat this as a lower bound.

Has student use of ChatGPT increased?

Yes. Pew Research Center found the share of US teens using ChatGPT for schoolwork doubled from 13% in 2023 to 26% in 2024, and HEPI found any AI use among UK undergraduates rose from 66% in 2024 to 95% in 2026.

Do students think using ChatGPT to write essays is acceptable?

Mostly not. In Pew’s 2024 survey only 18% of US teens said using ChatGPT to write essays is acceptable and 42% said it is not, compared with 54% who said it is acceptable for researching new topics.

What do students mainly use AI for?

Support tasks dominate. In HEPI 2026, 61% used AI to explain concepts, 49% to summarise an article, 40% to suggest research ideas and 39% to structure their thoughts for assessed work.

Is Turnitin data a reliable measure of ChatGPT essays?

Not on its own. Turnitin reported that about 11% of more than 200 million papers had at least 20% AI writing flagged, but these are classifier results rather than confirmed cases, so they can include misjudged human writing and miss edited AI text.

Can I use these statistics in my thesis literature review?

Yes, if you cite the original report, give the population and fieldwork date, and quote the question wording. Avoid combining figures from different surveys into a single prevalence claim.

Does using ChatGPT count as plagiarism in a thesis?

It depends on your institution’s policy and on what the tool did. Submitting AI-generated text as your own is restricted almost everywhere, while uses such as explaining concepts are often allowed if declared, so check your programme’s rules before you start.