Is using AI to revise cheating?
No. Revising with AI is not cheating, because revision is not assessed. What is assessed is the work you hand in and the paper you sit, and there the rule is simple: anything submitted under your name has to be yours. JCQ, which publishes joint guidance for the main UK exam boards, treats AI-generated content passed off as a student's own work in coursework or non-exam assessment as malpractice, and expects AI use to be acknowledged where it happens. Using a tutor, human or otherwise, to understand a topic has never been against any of those rules.
This question gets asked in two very different tones of voice. A student asks it quietly, half expecting to be told off. A parent asks it because they have read a headline and want to know whether their child is about to be disqualified. Both deserve a straight answer, and the straight answer is that almost everything people mean by revising with AI is fine, and one specific thing is not.
Revision is not assessed
Start with the obvious thing that somehow gets lost. No exam board has ever cared how you learned something. There is no rule against being taught by a private tutor, or by your older sister, or by a YouTube channel, or by reading the textbook four times. The mark is for what you can produce in the exam. How the understanding got into your head is not part of the assessment.
That is the whole answer for the ordinary case. If you sat down last night and asked something to explain oxidation states until it clicked, you did not cheat. You did the thing every student who can afford a tutor has always done, using a tool that costs less.
The line is authorship, not assistance
Where it changes is the moment something you did not write goes in with your name on it. That is not a new rule invented for AI. Copying a friend's essay, buying one, or lifting paragraphs from a website were all malpractice long before any of this existed, for the same reason: the qualification certifies that you can do the work, so work you did not do makes the certificate false.
JCQ, the body through which AQA, Edexcel, OCR, WJEC and the others publish joint guidance, has applied that principle to AI directly. Its guidance treats AI-generated content submitted as a student's own work as malpractice, and expects use of AI to be acknowledged where it has contributed to something being assessed. Centres are also expected to take reasonable steps to be confident that assessed work is the student's own.
Where the risk actually is
Almost all of the exposure sits in one place: work done outside an exam hall that carries marks. Depending on your subject that is coursework, a non-exam assessment, an EPQ, a practical write-up, a language piece, a programming project. These are the components where you are unsupervised and the output is graded, so they are the components the guidance is written about.
- Written exams: closed conditions, no devices, so the question does not arise.
- Coursework and NEA: the sharp end. Anything drafted or generated for you belongs to whatever produced it, not to you.
- EPQ: the process is part of the mark, and your production log is meant to record what you actually did.
- Homework and practice essays: usually not certificated, but your school's own policy still applies, and a practice essay you did not write teaches you nothing.
- Mock exams: sat under exam conditions, so treat them as exams.
Note that last one about homework. A practice essay written for you is not a rule problem so much as a self-inflicted wound. The point of writing it is to find out what you cannot yet do while there is still time to fix it. Outsourcing that hides the gap until the day it costs you.
Your centre's policy sits on top
Exam board guidance is the floor, not the ceiling. Schools and colleges are free to be stricter, and many are, particularly about work submitted through their own systems. Some require a declaration on anything assessed. Some ban AI for specific subjects outright. If your school's rule is narrower than JCQ's, your school's rule is the one you will be dealt with under.
If you are unsure, the least costly thing you can possibly do is ask your teacher before you hand something in. Nobody has ever been penalised for checking.
What using it properly looks like
The useful test is whether the work in front of you got harder or easier. Revision that is working is uncomfortable: you are retrieving, attempting, failing and correcting. If a tool has made the retrieval unnecessary, it has taken the learning with it.
- Get it explained. Ask for the topic in a different way from the one that did not land, and keep going until you can say it back without looking.
- Attempt the question yourself first, on paper, badly if necessary.
- Then ask where your attempt went wrong, and why. That is the sentence worth having.
- Get quizzed on it, with the answers hidden until you have committed to one.
- Write the essay yourself. If you want feedback on it, ask for feedback, not for a rewrite.
None of that produces a single word you would have to declare, because none of it produces a word you did not write.
The version of this that is genuinely a problem
It is worth naming honestly. A student who pastes a coursework title into a chatbot, lightly edits what comes back and submits it has committed malpractice, and if it is detected the consequence can reach the qualification itself rather than the piece of work. That is a real risk and it is not worth taking for one component.
It is also, quietly, the worst possible trade. The essay you did not write is the topic you will not be able to answer on in June, when there is nothing to paste into anything.
What to take from this
- Revision is not assessment. Nobody marks how you learned something, only what you can do in the exam.
- The line is authorship. Understanding an idea from AI is fine; submitting text or code it produced as your own is malpractice.
- Coursework and NEA are where the risk sits, because those are the marks awarded for work done outside an exam hall.
- Written exams are closed conditions, so the question of AI in them does not arise.
- Your school or college may have its own policy that goes further than the exam boards. That policy is the one you are held to.
- Ask AI to explain, quiz you and mark your attempt. Do not ask it to write the thing you are about to hand in.
Questions people also ask
There is nothing to find out. Revision is not submitted and not assessed, so there is no rule being broken and nothing to detect. The question only arises for work you hand in.
Only within what your exam board and your centre allow, and only where you acknowledge it. The safe position is that every word of assessed work should be yours. If you want to use AI around it, ask your teacher first and follow whatever declaration process your centre uses.
They are not, in either direction, which is one reason centres tend to rely on knowing a student's usual work and on the drafting process rather than on a score from a tool. That cuts both ways: do not assume you would get away with it, and do not assume you are safe because a detector says so.
Not in principle. A tutor who explains a topic is fine. A tutor who writes your coursework is malpractice, and has always been. The medium changed; the line did not.
Fine. Neither is assessed, and both are organisational rather than authorial. Just check the content against your actual specification, because a plausible-sounding revision list can quietly include topics your board does not examine.
JCQ publishes AI guidance for centres and candidates, and each exam board restates it for its own qualifications. Your school will also have its own policy. Read the current versions rather than any summary, including this one, because this area has been updated repeatedly.