94 per cent of full-time UK undergraduates say they use generative AI to help with assessed work, and 95 per cent use AI in at least one way, according to the HEPI/Kortext Student Generative AI Survey 2026 (Report 199, published 12 March 2026; fieldwork by Savanta, December 2025; 1,054 respondents).
Two years ago the same survey series was asking whether students had tried these tools. The 2026 question is no longer whether but how — and the most consequential number in the report is not the 94 per cent but the smaller one underneath it: the share of students putting AI-generated text directly into assessed work has quadrupled since 2024. Every figure below is from the published survey unless attributed otherwise, and each is written so it can be quoted alone.
The headline figures
| Measure | Value | Source and year |
|---|---|---|
| Use AI in at least one way | 95% | HEPI/Kortext Survey 2026 |
| Use generative AI to help with assessed work | 94% | HEPI/Kortext 2026 |
| Include AI-generated text directly in assessed work | 12% | HEPI/Kortext 2026 |
| — same measure, 2025 | 8% | HEPI/Kortext 2026 (trend) |
| — same measure, 2024 | 3% | HEPI/Kortext 2026 (trend) |
| Say assessment has changed significantly in response to AI | 65% | HEPI/Kortext 2026 |
| Feel encouraged by their institution to use AI | 36% | HEPI/Kortext 2026 |
| Say their institution provides AI tools | 38% | HEPI/Kortext 2026 |
| Believe AI has improved their student experience | 49% | HEPI/Kortext 2026 |
| Believe AI skills are essential to thrive today | 68% | HEPI/Kortext 2026 |
| Use AI for companionship, advice or loneliness | ~15% | HEPI/Kortext 2026 |
What does the survey actually measure?
The population is full-time UK undergraduates — 1,054 of them, surveyed by Savanta in December 2025 and reported by the Higher Education Policy Institute with Kortext in March 2026. That scope matters when you quote it: the figures describe undergraduates, not postgraduates, and self-reported behaviour, not observed behaviour. Self-report cuts both ways on a topic with a misconduct shadow — some respondents may understate uses they think are banned, others may count a grammar checker as “AI” — so treat the numbers as the best available estimate of a population that has every reason to be coy, not as an audit.
The number that matters: 3% → 8% → 12%
“Using AI to help with assessed work” spans everything from asking for an explanation of a concept to summarising an article — most of it uncontroversial and much of it officially encouraged. Including AI-generated text directly in submitted work is a different category, and it is the one moving fastest: 3 per cent in 2024, 8 per cent in 2025, 12 per cent in 2026. Whether any given instance is misconduct depends entirely on the module’s rules — some assessments now explicitly permit disclosed AI drafting — but the trend line explains why 65 per cent of students say assessment itself has changed significantly in response to AI. Universities are redesigning around exactly this behaviour.
The support gap
Set the adoption figures against the institutional ones and the tension is obvious: 94 per cent of students use generative AI on assessed work, but only 36 per cent feel encouraged by their institution to use it and only 38 per cent say their institution provides AI tools. Fewer than half — 48 per cent — feel teaching staff help them develop the AI skills they will need for their careers. In other words, the behaviour is universal while the guidance, provision and training remain minority experiences. For students the practical consequence is that the rules governing your AI use are local and often new: the module handbook and each assignment brief are where your actual permissions live, and they can differ between two modules on the same course.
Reading survey data critically — a dissertation skill in itself
If you plan to cite these figures in your own work, practise on them the scrutiny your markers will practise on you. Check the population before generalising: a finding about full-time undergraduates says nothing direct about part-time or postgraduate students. Check the instrument: “use AI” is whatever the questionnaire defined it to be, and comparisons across different surveys with different wordings are not comparisons. Check the trend’s provenance: the 3–8–12 series is quotable because it comes from the same survey series asking the same question, which is exactly what makes it stronger evidence than two unrelated polls. And note what is not measured — detection, misconduct outcomes, actual marks — so you do not smuggle claims the data cannot carry. That checklist is transferable to every statistic in your literature review.
How to use AI on a dissertation without gambling your degree
The survey describes what students do; it says nothing about what your department permits, and that distinction is the whole game. Three habits keep you on the right side of it. Read the assessment brief’s AI statement before touching a tool, and where none exists, ask — in writing. Keep the authorship yours: tools that structure, check and format leave you the author; pasting generated text into an essay that claims to be your own writing is the behaviour the 12 per cent figure tracks, and it is the one that ends up at misconduct panels where rules forbid it. And disclose what your department asks you to disclose — a use you can state openly is a use you have already vetted. The full method is in our guide to writing a literature review with AI, honestly, and the authorship line — why similarity software cannot settle it in either direction — is unpacked in our plagiarism checker comparison.
Used that way, AI assistance and academic credit point the same direction rather than against each other. Tesify is built for exactly that division of labour — structure, formatting and bibliography handled, every word still written by you.
Where AI shows up beyond coursework
Two findings worth quoting because they widen the picture. Around 15 per cent of students report using AI for companionship, advice or to address loneliness — a wellbeing datum hiding in an education survey. And 49 per cent believe AI has improved their student experience overall, against a landscape the report describes as polarised, with students split roughly evenly on several of the big questions. The average student is not an enthusiast or a refusenik; they are a pragmatist using the tools available for the work in front of them, with or without institutional blessing.
How to cite these figures
Attribute to the Student Generative AI Survey 2026, Higher Education Policy Institute and Kortext (HEPI Report 199, published 12 March 2026), fieldwork by Savanta in December 2025 with 1,054 full-time UK undergraduate respondents. Name the population when quoting — “of full-time UK undergraduates surveyed” — and use the report’s own trend figures for prior years rather than splicing numbers from different survey series, whose question wordings differ. Degree outcomes data, if you need it alongside, is covered in our degree classification statistics roundup.
Frequently asked questions
What percentage of UK students use AI in their studies?
95 per cent of full-time UK undergraduates report using AI in at least one way, and 94 per cent use generative AI to help with assessed work specifically (HEPI/Kortext, 2026).
How many students submit AI-written text?
12 per cent report including AI-generated text directly in assessed work in the 2026 survey — up from 8 per cent in 2025 and 3 per cent in 2024. Whether that constitutes misconduct depends on each assessment’s rules.
Is using AI for university work cheating?
Not inherently — most reported uses are explanation, summarising and study support, and many departments explicitly permit disclosed AI assistance. It becomes misconduct where use breaks the assessment’s stated rules or misrepresents authorship. The brief for each assignment is the authority.
Do universities provide AI tools to students?
Mostly not yet: only 38 per cent of students say their institution provides AI tools, and only 36 per cent feel encouraged to use AI (HEPI/Kortext, 2026).
Has assessment really changed because of AI?
65 per cent of students say assessment has changed significantly in response to AI (2026). The direction of change varies — more in-person and oral assessment in some departments, explicit AI-permitted tasks in others.
Who conducted this survey and how big is it?
The Higher Education Policy Institute with Kortext, fieldwork by Savanta in December 2025, 1,054 full-time UK undergraduates. It is the fourth wave of the series, which is what makes its trend lines quotable.
Does the survey cover postgraduate or part-time students?
No — the sample is full-time undergraduates. Quoting the figures for postgraduates or the whole student body misstates the population.
Are students confident AI skills matter for their careers?
68 per cent believe AI skills are essential to thrive today, while fewer than half feel teaching staff help them develop those skills — one of the report’s clearest gaps.
Can universities detect AI-generated writing?
AI detection is probabilistic and contested, and no figure in this survey measures detection. Universities increasingly assess authorship through assessment design — orals, drafts, supervised work — rather than relying on detectors alone.
Where can I read the full report?
HEPI publishes the Student Generative AI Survey 2026 as Report 199 on hepi.ac.uk, free to download, with the questionnaire wording and full tables — the right source to cite in academic work.