Accused of Using AI at University: A UK Step-by-Step Guide

An email arrives: your work has been flagged for possible AI use, and there's a meeting. Whether you used AI or not, what you do next matters far more than the score that started it. This guide sets out the UK process, the evidence that actually persuades panels, and a template you can adapt — calmly and without legalese.

It is written for the accused student. If you're an educator running the review, read our fair-review guide instead.

First 48 hours: do these five things

1. Read the allegation letter carefully. Note the exact regulation cited, the deadline to respond, and what evidence the university says it has. Do not reply in anger, and do not reply immediately.
2. Preserve everything — now. Do not "tidy up" any file. Version history, drafts, notes, search history and reference exports are your case; editing anything after the allegation can look like concealment.
3. Contact your students' union advice service. Most UK SUs run a free, confidential academic advice service that has seen dozens of these cases and can accompany you to meetings. This is the single most under-used step.
4. Check your university's own AI-detection position. If your university has publicly opted out of Turnitin's AI indicator or cautioned against detector reliance — many have — that context belongs in your response.
5. Tell the truth to yourself first. The strategy below is for students who did not break the rules, and for students whose permitted use was misread. If you did use AI beyond what was allowed, honesty plus mitigation is consistently the least damaging route — SU advisers will confirm this.

Understand what the university must show

UK universities run academic-misconduct cases under their own regulations, but the Office of the Independent Adjudicator's Good Practice Framework for disciplinary procedures sets the expected shape: a clear allegation, disclosed evidence, a genuine opportunity to respond, and a decision on the balance of probabilities by people who considered your side.

Two things follow for AI cases:

• A detector score alone is thin evidence. Detector vendors themselves say scores shouldn't decide outcomes, and detector results are classifications, not recovered histories. Panels know this — which is why most cases turn on the surrounding evidence: your drafts, your explanation, your ability to discuss the work.
• You are answering a specific allegation, not defending your character. Keep everything you write and say tied to the assignment, the rules that applied to it, and your writing process.

Gather your evidence

Bring documents, not assertions. In rough order of persuasive power:

• Version history (Word/OneDrive, Google Docs) showing the document growing over sessions. Our prove-you-didn't-use-AI guide explains how to export it.
• Drafts and outlines — earlier files, plans, mind-maps, supervision or seminar notes.
• Research trail — library history, saved sources, annotated PDFs, reference-manager entries.
• Your ability to discuss the work — re-read your submission and sources; being able to explain your argument and choices is evidence in itself.
• Context for the flag — quotations, templated sections and formal register commonly trigger false positives; non-native English writers are disproportionately flagged, a point HEPI has documented for UK universities.
• An independent, dated report of your own text — optional, but a sentence-level report you commission yourself (for example a Genutext scan emailed to you) shows which passages carry signals and demonstrates you engaged seriously; present it as context, never as counter-"proof".

What to say — and what not to say

Say:

• "Here is how this piece was written" — a short chronological account.
• "Here is the evidence of that process" — attach it.
• "Here is what I understood the module rules to permit" — quote the assessment brief.
• If you used permitted tools (spellcheck, Grammarly-style polish, translation help): name them plainly and say where.

Avoid:

• Arguing the detector's accuracy in the abstract as your main defence — it invites a technical debate you don't need to win. Your process evidence is the affirmative case; the detector's limits are context.
• "I can't explain it" — always offer your best account.
• Rewriting or re-running your work through tools after the allegation.
• Missing the deadline — ask for an extension rather than letting it pass.

A template response letter

Adapt freely; keep it under a page. Replace bracketed text.

Dear [name],

Thank you for your letter of [date] regarding [module/assignment]. I understand the concern raised is [state the allegation in one sentence]. I did not use AI tools to write this work [or: my use of AI was limited to (name tool and purpose), which I understood to be permitted under (brief/handbook reference)].

I have attached evidence of my writing process: [list — e.g. document version history exported from (platform), two earlier drafts, my outline dated (date), and my reference-manager library]. I would welcome the opportunity to discuss the work and my sources at the meeting.

I would also ask that the panel consider the flagged report in light of [university's/vendor's] own guidance that detector results should not be the sole basis for a finding[, and my position as a non-native English writer, a group documented to receive disproportionate false flags].

I am keen to resolve this and will make myself available at your convenience. My students' union adviser [name, if applicable] will accompany me.

Yours sincerely, [name, student number]

The meeting and afterwards

• Take someone. An SU adviser or permitted companion changes the dynamic and takes notes.
• Answer questions about the work directly. Panels routinely ask you to explain passages, define terms you used, or describe a source. This is your strongest ground — treat it as a viva, not an interrogation.
• Ask what happens next and when. Get the outcome and reasoning in writing.

If the outcome goes against you

Every university has an internal appeal route — usually on grounds of procedural irregularity, new evidence, or an unreasonable decision. Deadlines are short (often 10–15 working days). Exhaust it first; then, in England and Wales, you can take the case to the OIA within 12 months of the final internal decision (Scotland and Northern Ireland have equivalent ombuds routes). SU advisers handle appeals weekly — use them again.

Frequently asked questions

Can a university prove I used AI?

Rarely in a forensic sense — which is why cases are decided on the balance of probabilities using all the evidence: the flagged report, your work's characteristics, your process evidence and your account. Strong process evidence usually decides it.

What should I say when accused of using AI?

Give a short, factual account of how you wrote the piece, attach process evidence, state what the module rules permitted, and offer to discuss the work. Avoid making the detector's general unreliability your only argument.

What happens if you get accused of using AI at university?

Typically: a written allegation, an invitation to respond or attend a meeting, a decision (no case, a penalty, or escalation to a panel), and an appeal route. The OIA's framework expects you to see the evidence and be heard before any finding.

Should I admit to using AI if I did?

If your use broke the rules, an honest account with context and mitigation is consistently the least damaging route, and advisers will tell you the same. Denials that unravel make outcomes worse.

Can I be accused because of a false positive?

Yes — human writing gets flagged, formal and non-native prose most often. That is exactly why your drafts and version history matter more than the score.

Sources and further reading

• OIA: Good Practice Framework — academic disciplinary procedures
• OIA: Casework note — complaints relating to academic misconduct
• HEPI: Catching the wrong students — AI detection and international students (20 July 2026)
• Genutext: Can an AI detector score be used as evidence? UK misconduct rules
• Genutext: How to prove you didn't use AI
• Genutext: UK University AI-Detection Policy Tracker

This guide is general information about UK higher-education procedures, not legal advice. Genutext is independent and not affiliated with any university, students' union or the OIA. Sources current as of 23 August 2026.