Plagiarism and AI at UCL: the investigatory viva explained

The short answer

At UCL, a doubt about who wrote your work is resolved by talking to you, not by a number. The university runs an investigatory viva when authorship is in question, and its student guidance says explicitly that markers do not run GenAI detectors.

That changes what you should prepare. There is no score to dispute and no threshold to stay under. What matters is whether you can talk about your own argument — which is a much fairer test, and a much easier one if you actually wrote the thing.

What an investigatory viva is

It is a meeting about your own work, and it is written into the procedure. UCL uses it precisely where a tool cannot help:

“Refer the case to an Investigatory Viva for consideration under Stage 2 where authorship is in question (usually Contract Cheating or Misuse of GenAI).”

“In cases where authorship is in question (normally allegations of Contract Cheating and Misuse of GenAI), the Department/Faculty will organise an Investigatory Viva and the Secretary will prepare the paperwork and invite the student to attend.”

The student-facing guidance adds the trigger, and a reassurance worth reading twice: “If teaching staff suspect that you are trying to pass off GenAI output as your own work then the regulations in the Academic Manual […] may apply. You may be invited to an investigatory viva to probe the authorship of your work.”

So how do you prepare? Not by rehearsing a defence, but by being able to do what any author can do: say why you framed the question that way, what you read that changed your mind, which sources you rejected and why, and what you would do differently. If you have kept drafts and dated version history, bring them. We set out the practical steps in how to prove you wrote it yourself and in what to do if you are accused of using AI.

If a detector score is what started this, our guide to an AI detector flagging your thesis sets out what that number does and does not establish on its own.

Why there is no score to argue about

UCL’s position is on its own student assessment page, in one sentence:

“UCL does not use GenAI detectors when marking. Instead, if a marker has any concerns about your work, they may speak to you to get more information. You should not automatically assume this means that they suspect you of academic misconduct.”

We flag one thing honestly: that page carries no publication date and names no office on its face. It sits inside UCL’s Assessment Success Guide and cross-refers to the Academic Manual, so it is the student-facing expression of the same regime — but if you need to rely on it formally, ask your department to confirm it in writing.

The last sentence is the one students miss. A marker asking you about your work is not an accusation. At UCL it is the normal first step, and answering it well usually ends the matter.

This is also why we say the same thing about our own tool as UCL says about everyone’s: a detector produces a probability with an error rate, not proof. Our published rates are measured on English and German corpora, and we would not want a marker to treat any score — ours included — as an answer to a question that a five-minute conversation answers better. More on what these numbers can and cannot do in how accurate AI detectors really are.

The rule that protects non-native speakers

If you are writing your degree in a second language, this clause is worth knowing by heart. It appears in the procedure itself:

“The difference between a student’s spoken and written English alone are not necessarily a cause for concern that would warrant investigation.”

In other words, writing more formally than you speak is not evidence of anything. That is exactly the pattern that gets non-native speakers wrongly suspected — careful, textbook-correct prose that reads as “too polished” — and UCL has written down that it is not, by itself, a reason to investigate. We wrote about the wider problem in AI detectors and non-native speakers.

What counts as misuse of GenAI

UCL lists misuse of GenAI as its own category of academic misconduct, alongside plagiarism and contract cheating:

“Unauthorised and/or unacknowledged use of Generative Artificial Intelligence (Misuse of GenAI): defined as the use of assistive tools beyond the remit of the assessment brief, or to create false data, text, or references.”

Read the definition closely, because it has two halves and the second one catches people out. Beyond the remit of the brief means your module leader sets the boundary — the same tool can be fine in one module and misconduct in another. False data, text, or references is the hallucinated-citation problem, and it is the most common way a student who was not trying to cheat ends up in a procedure. Check every reference exists in a real database before you submit; see verifying AI citations.

Two neighbouring categories are worth knowing. Self-plagiarism — resubmitting your own previously assessed work — is listed as misconduct, which surprises people reusing part of an earlier essay. And contract cheating carries a warning few universities print: “Students should be aware that it is a criminal offence to advertise or provide Contract Cheating services, and UCL would normally report this to the Police.”

UCL also defines a lighter category, and it is the one most first-year referencing mistakes belong in: “Poor Academic Practice refers to low level academic integrity infringements that are clearly judged to be a result of a lack of understanding on the part of the student (e.g. inadequate referencing or poor paraphrasing).”

What UCL explicitly allows

This is the clearest carve-out we have found at any university, and it answers the question most students actually have. Quoted in full, including its limit:

“Language and Writing review: defined as having a third-party or software check areas of academic writing such as structure, fluency, presentation, grammar, spelling, punctuation, and language translation. However, this may be considered Academic Misconduct if substantive changes to content have been made by the reviewer or software or at their recommendation, which would suggest that the reviewer had either produced or determined the substantive content of the submission, or, in cases of language translation, if the student is being assessed on their ability to translate or use a language other than English.”

The line is between how it reads and what it says. Fixing grammar, tightening a sentence, catching a misplaced comma: fine. Having software decide what your argument is: not fine. If a suggestion changes your meaning rather than your phrasing, it has crossed over.

UCL’s guidance draws the same line for GenAI generally: “Using GenAI tools to check your grammar or spelling is usually appropriate if the tool is not rewriting a large proportion of your work.” And the hard stop: “It is not acceptable to use GenAI tools to write your entire assessment and present this as your own work.”

Worth knowing too: UCL funds a writing-feedback service for every student, and it is itself GenAI-based — “UCL offers a service called Studiosity to support students with their academic writing and referencing skills. The platform uses GenAI and is available 24/7 to all current students, at all levels of study.” It also names Microsoft Copilot as “our approved GenAI tool because of its enhanced data protection”. If you are going to use something, using the one the university sanctioned is the easier conversation.

Acknowledging AI use

There is no single university-wide declaration form with prescribed wording — we looked, and none is published. The duty sits at module level:

“Module leaders are responsible for providing accurate and up-to-date information to students about their assessments. This includes information about appropriate referencing, clear instructions about collaborative working, the scope of permissible use of Generative Artificial Intelligence (GenAI), and a link to these Student Academic Misconduct Procedures.”

And the default, when your brief says nothing: “If no explicit instructions are given, the default for assessments should be that you use GenAI in an assistive role while making sure that the final submission is still substantially your own work.” Note that this is more permissive than several other universities, where silence means no.

The duty to acknowledge is stated simply: “You should acknowledge GenAI use where it has been used to assist in the process of creating your work.” How to word that in three to five lines is in our AI disclosure statement guide.

The penalty table

UCL publishes its tariff, which not every university does, and it escalates by who is deciding.

Department or faculty may reduce the component mark by 10 percentage points or one letter grade, cap the component at the pass threshold, or give the component 0.00%/Grade F. For exam misconduct only, the mark can stand with a formal written warning.

An Investigatory Viva Panel or Departmental Panel can do any of the above, plus cap the whole module at the pass threshold or give the module 1.00%/Grade E.

The UCL Academic Misconduct Panel can do any of the above, plus:

• Give the module 1.00%/Grade E and annotate your official transcript to reflect the penalty.
• Give the module 0.00%/Grade F.
• Suspend you from UCL for the remainder of the academic year.
• Exclude you from UCL with no re-enrolment on any other UCL programme — with an interim qualification still possible if you are eligible.
• Exclude you with no re-enrolment and no qualification, even if otherwise eligible.

And after graduation, verbatim:

“Where a case of academic misconduct is upheld against a graduate, UCL reserves the right to revoke the award and apply the appropriate penalty. This may include the decision not to reissue the award in cases of serious academic misconduct.”

One rule about how penalties are set is easy to misread, so here it is plainly. Mitigating circumstances do not affect whether misconduct happened — “A decision about whether academic misconduct has occurred will not take mitigating factors, including extenuating circumstances, into account” — but they doaffect the penalty, which “must take into account any available mitigating factors, as well as the impact such penalty will have on the individual student’s academic progression”. If you had a hard year, that belongs in the penalty stage, not the finding stage.

Process, proof and appeal

Three stages: an initial departmental investigation, then a panel, then a paper-based appeal review. Contract cheating and research misconduct by taught students skip straight to the university panel.

The standard of proof is written down, and it is not certainty: “In considering allegations of Academic Misconduct adjudicators should apply ‘on the balance of probabilities’ as the standard of proof.” More likely than not, in other words. That is precisely why turning up with concrete material — drafts, notes, version history — matters more than a denial.

Classification matters for what you face. A first offence of plagiarism, self-plagiarism, misuse of GenAI, collusion, falsification or exam misconduct is normally a minor case, as is a second, and as is any case “where there has been minimal to no academic advantage gained”. Cases become major at two or more upheld offences on record, or for contract cheating, impersonation, research misconduct, or fraudulent documentation affecting progression or award.

Appeals must be lodged within 10 working days. And once UCL’s internal process is exhausted, there is an external route: “When a student has received a Completion of Procedures letter […] they have the right to submit a complaint for review to the Office of the Independent Adjudicator for Higher Education (OIA)”, within twelve months.

UCL does not publish annual counts of misconduct cases, so nobody can honestly tell you how often any of this happens.

Before you submit

• Read the assessment brief for the GenAI scope. That is where the boundary lives at UCL, not in a central rule.
• Keep drafts and version history from day one. The process ends in a conversation about your work; evidence of process is what makes that conversation easy.
• Check every citation exists. Fabricated references are named in the definition of misuse.
• Keep edits at the level of phrasing, not content. That is the line UCL actually draws.
• If a marker asks you about your work, treat it as a question, not a charge. UCL says so itself.

If you want a read on your own text before you hand it in, our AI checkis free up to 1,500 words, with no account and no name. It marks filler, hedging and tired transitions in your own sentences and suggests plainer wording — the “how it reads” side of the line above, not the “what it says” side. Our published error rates are measured on English and German corpora, and no detector, ours included, produces proof.

Sources

• UCL: Academic Manual, Chapter 6, Section 9 — Student Academic Misconduct Procedure — the page carries no version number or approval date on its face.
• UCL: Engaging with Generative AI in your education and assessment — the statement that UCL does not use GenAI detectors; undated.
• UCL: Annex 6.9.8 — Penalty Table for Student Academic Misconduct.
• UCL: Language and Writing Review guidance.

This page is orientation, not legal advice. What binds you is your assessment brief and the UCL Academic Manual.