UK procedure
UK Academic-Misconduct Procedures and AI Evidence
Map the common stages of a UK case while recognising that institutions differ on detector use, vivas, evidence and decision routes.
On this page
There is no single UK-wide university procedure for suspected unauthorised AI use. Each provider defines misconduct, evidence, decision stages and appeals in its own regulations. Some institutions use AI-detection reports cautiously; others do not use them at all.
Across those differences, fair procedures usually share a core: a clear rule, a specific allegation, disclosure of relevant evidence, an opportunity to respond, an evidence-based decision and a route to review or appeal.
This guide is general information, not legal advice. Students and staff should read the regulations and assessment instructions that apply to the specific case.
Who sets the rule
Unauthorised AI use becomes misconduct through the applicable institutional and assessment rules—not through the detector's label.
The rule should make clear whether students may use generative AI for brainstorming, research support, editing, translation, coding or drafting, and whether disclosure is required. A blanket institution statement may be supplemented by stricter or more permissive module instructions.
The OIA's good-practice framework says providers should define types of academic misconduct and communicate them clearly. That matters with AI because “use” ranges from a spell correction to generation of assessed content.
A typical case pathway
Although labels differ, a process may include:
- Initial concern. A marker identifies a specific issue in submitted work.
- Preliminary review. An authorised staff member checks the rule, evidence and whether the concern can be resolved or should proceed.
- Written notification. The student receives the alleged offence, reasons and available supporting evidence.
- Student response. The student can supply an explanation, drafts, sources and other relevant material.
- Investigation or meeting. Straightforward cases may be resolved early; complex or serious cases may go to an academic-misconduct officer, viva or panel.
- Decision. The decision-maker applies the published standard of proof and gives reasons.
- Penalty, if a breach is found. The sanction should follow the provider's tariff and consideration of relevant factors.
- Review or appeal. The written outcome explains the available route and deadline.
The OIA says students should be told in writing what offence is suspected and why, receive available supporting evidence and have an initial opportunity to respond.
Try the text in context
Run a free AI-writing signal check
Paste 300 characters to 350 words. The sample is analysed for a first-pass signal and then discarded.
How AI detector evidence fits
An AI report may help identify passages for review, but it does not determine whether a policy was broken.
The report's value depends on:
- whether the tool was approved for the material;
- supported language, length and genre;
- the complete report and highlighted passages;
- a documented explanation of the score;
- known false-positive and false-negative limitations;
- the report date and model version; and
- independent evidence of the writing process.
The OIA treats interpretation of academic-misconduct detection reports as academic judgement and recommends sharing the academic analysis with the student as well as the report. Turnitin similarly says its model may misclassify text and should not be used alone for adverse action.
Read the operational checklist in what to do when a student disputes an AI result.
Burden and standard of proof
Policies should state who must establish the allegation and to what standard. Many UK providers use the civil standard—often described as the balance of probabilities—but the exact wording and evidence rules must be checked locally.
The University of Bristol, for example, states that academic misconduct is found where the evidence makes the offence more likely than not. That does not turn a detector score above 50% into proof. A statistical percentage is not the legal or procedural standard; the decision-maker weighs the complete evidence.
A student may provide drafts or version history, but that does not mean the burden automatically shifts to the student to disprove an algorithm.
Different UK approaches
Current institutional examples show why generalisations are unsafe:
- University of Glasgow: its 2026 guidance says the university does not employ AI-detection tools and investigations should never rely on such software.
- UCL: its procedure allows an investigatory viva as an initial step for suspected unauthorised or unacknowledged generative-AI use where relevant.
- University of Bristol: its regulations describe panel routes and the balance-of-probabilities standard.
- University of Cambridge: institutional pages direct staff to specific guidance on investigating suspected AI-related misconduct.
These are examples of local policy, not a league table. Always use the current version of the provider's own regulations.
Privacy and automated decisions
Student work, detector inferences and misconduct case records can be personal data. Institutions should use approved systems, give appropriate privacy information, define retention and access, and avoid uploading work to unapproved third-party tools.
UK data-protection guidance treats predictions or inferences about a person as personal data. Where a significant decision is made solely by automated processing, specific safeguards apply, including information, a route to representations and meaningful human intervention. Even where a human makes the decision, token approval of an algorithm is not a substitute for genuine review.
Staff should follow the institution's data-protection impact assessment, procurement and records policies rather than making an individual decision to upload student work.
Frequently asked questions
Can a UK university punish a student based only on Turnitin AI detection?
The applicable regulations govern the case, but Turnitin itself says the report should not be the sole basis for adverse action. OIA good practice expects evidence-based judgement and a fair opportunity to respond.
Does every UK university use the balance of probabilities?
No universal assumption should be made. Many do, but students and staff must check the current provider regulations for the stated burden and standard.
Can a student ask to see the AI report?
Fair procedure normally includes disclosure of the allegation and relevant supporting evidence. The precise access route depends on the provider's regulations and data-protection process.
Is an investigatory viva compulsory?
No. Some providers use vivas in particular cases; others use meetings, written responses or panels. The published local procedure controls.