AI in education: 92% of students use it, unis lag
92% of UK undergrads use AI. Only 36% got trained.

92% of UK undergraduates were using AI tools by the end of 2024. 88% were using them on assessed work.
and only 36% said their university had actually trained them to use it.
that's the state of AI in education going into 2026. students moved fast. universities are still writing the rulebook (and in some cases tearing pages out of it).
How students actually use AI at university
the clearest UK numbers are still from HEPI's Student Generative AI Survey 2025, which polled 1,041 full-time undergrads in December 2024:
→ AI use jumped from 66% to 92% in a single year → use on assessments jumped from 53% to 88% → the top uses were explaining concepts, summarising articles and suggesting research ideas → 42% said staff were well equipped to help with AI, up from 18%
the US picture looks similar. Inside Higher Ed's Student Voice survey of 1,047 students at 166 institutions, run in July 2025, found:
- 85% used generative AI for coursework in the past year
- the most common use was brainstorming (55%), then asking it questions like a tutor (50%)
- 55% of those users said the effect on their learning was mixed: useful, but it made them think less deeply
- only 21% wanted universities to lean on AI detection software
then a January 2026 Honorlock survey of 1,005 US students adds a twist. 56% say an instructor has REQUIRED them to use AI on at least one assignment.
(worth knowing: Honorlock sells exam proctoring software, so read its numbers with that in mind.)
so this isn't just students sneaking ChatGPT in. some lecturers are assigning it.
Where university AI policies are falling behind
this is the messy bit. the rules haven't settled.
the clearest example this month: Washington State University cancelled its Turnitin AI detection contract in February.
the provost's memo did the maths for everyone:
- Turnitin analysed 148,547 assessments at WSU in autumn 2024
- even a 1% false positive rate (work wrongly flagged as AI) means almost 1,485 of them
- so plagiarism checking stays, AI detection goes
put the whole picture side by side:
→ students: using AI on nearly everything → lecturers: some banning it, some requiring it → universities: switching off the tools meant to police it
and you sit in the middle, getting "use it" from one module and "that's misconduct" from the next.
What uni won't teach you about AI
most courses still treat AI as either a cheating problem or a productivity hack. both miss the point.
what hardly anyone teaches:
- how to check an AI answer against a primary source, fast
- when NOT to use it, because sometimes the struggle IS the learning
- how to disclose it properly, so you never end up in a misconduct meeting
- how employers will judge it: 81% of students in the Honorlock survey feel confident using AI at work, yet fewer than 1 in 5 have taken a class on using AI professionally
(confidence and training are not the same thing.)
The critical-thinking move: 5 questions before you open the chatbot
our whole approach is to think critically about the tools, not just use them. (our guide to thinking critically about AI goes deeper.) for uni work, run these five:
- what is this assignment actually testing? if it's your reasoning, outsourcing the reasoning defeats the point
- what are the rules for THIS module? not the uni's general policy, the one in your module handbook. screenshot it
- can I explain every line I submit? if a tutor asked you in a viva (an oral exam), could you defend it with the chat closed?
- where did this claim come from? every fact and citation traced to a real source you've opened yourself
- am I faster, or just busier? over half of US student users said AI made them think less deeply. notice when that's you
What to do this month
- find your module rules. for every course this term, write down whether AI is allowed, banned or required, and how to disclose it
- run a verification drill. ask an AI to summarise a reading you've already done, then mark it. count what it got wrong or left out
- keep an AI log for one assignment: what you asked, what you kept, what you changed. it doubles as your disclosure note
- do one task the hard way first, then compare it with the AI version. that gap is your actual learning
we'll keep tracking how AI is reshaping the sector on our education hub.
so TLDR: almost every student uses AI now and universities are still deciding what counts as cheating, so build your own rules before someone builds them for you.
which of your modules has the most confusing AI rules right now?
Sources
- Student Generative AI Survey 2025Higher Education Policy Institute (HEPI) and Kortext · 26 FEB 2025
- How AI Is Changing, Not Killing, College (Student Voice survey)Inside Higher Ed · 29 AUG 2025
- Student AI SurveyHonorlock · 06 JAN 2026
- Cancellation of Turnitin AI detection software: memo to instructorsWashington State University, Office of the Provost · 11 FEB 2026
Get the next issue by email
theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.