Plagiarism and AI at Imperial: two definitions that do not match

The short answer

Imperial put AI into two different definitions in March 2023, and they do not say the same thing. One treats submitting AI-generated work as plagiarism outright; the other requires it to have been done without acknowledgement or authorisation.

That difference is not academic. The two sit on separate rows of a published penalty tariff, and the harsher row reaches the rescinding of a degree already awarded.

The two definitions

Both come from the Academic Misconduct Policy and Procedure, version 1.3, effective 15 March 2023 — the revision whose own change log records “updates to academic misconduct definition”.

Clause 14(i), plagiarism:

“The presentation of work for any type of assessment which contains, intentionally or unwittingly, the unacknowledged published or unpublished words, thoughts, judgements, ideas, structures or images of some other person or persons work, including work generated by artificial intelligence and/or machine learning […] This includes […] the commissioning, purchase, and submission of a piece of work, in whole or in part, as the student’s own or the submission of work which has been generated by an artificial intelligence or machine learning programme, including translation apps.”

Clause 14(v)(d), dishonest practice:

“The submission of work for assessment which has been generated through an artificial intelligence of translation programme without acknowledgement or authorisation.”

(The slip — “artificial intelligence oftranslation programme” — is in Imperial’s original.)

Note what both share: translation tools are named, twice. If you draft in another language and translate, Imperial treats that as being within scope, which matters a great deal for international students.

Why the mismatch matters

Read the two limbs side by side. The plagiarism limb says “the submission of work which has been generated by an artificial intelligence […] programme” — with no qualifier attached to that clause. The dishonest practice limb says the same thing but adds “without acknowledgement or authorisation”.

On a literal reading, acknowledging your AI use answers the dishonest practice limb but not obviously the plagiarism one. We are not going to pretend that is tidy, and we are not going to resolve it for Imperial — it is their text.

What we can tell you is the practical consequence, because the tariff treats the two categories differently and dishonest practice carries the harsher range. So which heading an allegation is brought under changes your exposure materially. If you are ever in this position, it is a fair and important question to ask: which limb am I charged under, and why that one?

The general definition, clause 12, sets the backdrop and does not mention AI at all: misconduct is “any activity, intentional or otherwise, that is likely to undermine the academic integrity of scholarship and research […] Academic misconduct can result from a deliberate act […] or may be committed unintentionally.”

The published tariff, and where it ends

Imperial publishes its penalty ladder as options a to k, with a tariff table mapping offence category and student level to a range. Panels may depart from it “but must provide a clear written rationale”.

The lower rungs are genuinely remedial. Penalty b is “re-submission required with correction to referencing/citation (not permitted to change substance of the work)”; penalty c marks the work “by removing any text to be considered plagiarised”. The policy notes these “would not normally be utilised by an Academic Misconduct panel except where there are extreme significant mitigating factors or where is it considered that the offence is due to poor academic practice.”

The upper rungs are where it becomes serious:

• h — zero for the module, no retake, no credit, no award: expulsion.
• i — zero for all assessments and modules, no re-entry, no credits, no award: expulsion.
• j — where an award has already been made: withdraw credit and “rescind the award of a degree/diploma/certificate of the College”, retaining other credits.
• k — zero for all assessment and credit, and rescind the award entirely.

Two things make that reach further than at most universities. The Academic Misconduct Panel imposes j and k itself — there is no referral upward required. And the policy expressly looks backwards: “An allegation of academic misconduct may be investigated at any point, whether or not a final mark has been agreed for the work in question, or if an award of the College has been made.”

Imperial also publishes what “moderate” and “severe” mean numerically: moderate is “between 15 and 50%” of the assessment plagiarised “with little attempt to appropriately reference”; severe is “more than 50% […] with little or no attempt at appropriate citation”. That is one of the few places in this series where a percentage appears in a rulebook rather than in a similarity report — and it is about how much of your work is affected, not about a Turnitin score.

Panels may also require “further training in Academic Integrity such as the ‘plagiarism awareness’ course”.

If you are a doctoral student

This is the most important line on the page for anyone writing a thesis at Imperial, and it is easy to miss:

“Potential academic misconduct in postgraduate research programmes will be dealt with under ordinance D17, The Investigation of Allegations of Research Misconduct.”

Everything else on this page — the tariff, the panel, the definitions — is the taught-programme regime. A PhD or MPhil thesis at Imperial goes down a different route entirely, under the research misconduct ordinance.

If you are a research student, do not plan around the policy described here. Ask your department for Ordinance D17 and read that instead.

A report alone is not an allegation

Imperial names Turnitin™ once, and only to impose a condition on using it:

“Where the evidence is provided from a plagiarism detection or originality checking service, such as Turnitin™, a commentary must be provided as part of the report to indicate the location and the level of the alleged plagiarism. They should also clearly indicate any areas that should not be considered, for example template wording, or terms of specific meaning/definition for which a match would be expected.”

So a raw score cannot start a case. A member of staff has to locate the alleged plagiarism, grade its level, and actively exclude the matches that were always going to appear — template wording, standard terminology, defined terms. That is a real procedural protection, and if an allegation arrives without it, its absence is worth pointing out.

On AI detection specifically, Imperial publishes no position at all. We checked the misconduct policy, the academic integrity document, the Generative AI Principles (last updated 5 March 2025) and the student AI and Study Guidance Hub, and none mentions detection tools in either direction.

Whatever any institution uses, the limit is the same and applies to us: our published error rates are measured on English and German corpora, and no detector produces proof. See what a Turnitin score actually means and how accurate AI detectors really are. If a case has started, how to prove you wrote it yourself and what to do when you are accused set out what to bring.

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.

Consent you gave at registration

Imperial’s position on the comparison database is short and worth knowing:

“The College may submit your work to an external plagiarism detection service, and by registering with the College you are automatically giving your consent for any of your work to be submitted to such a service.”

Consent is deemed by enrolment, and no opt-out is published — the opposite of Cambridge, which grants a removal right, and of Toronto and Waterloo, which offer alternatives.

One documentary point we think you should know. That clause sits in Examination and Assessments: Academic Integrity, Version 4, dated September 2018 and approved by Senate — a document that has not been revised sinceand therefore predates generative AI entirely. It contains no reference to AI at all. The retention period, and whether Imperial submissions join Turnitin’s cross-institution pool or an Imperial-only one, is not published.

Your department writes the AI rule

Imperial publishes no College-wide AI declaration wording. Its student checklist puts the duty on you to go and find the local rule, and it is the first item:

“Review your department’s current policy on using and disclosing genAI in academic work.”

Its Generative AI Principles — drafted by Imperial’s AI Futurists in consultation with the AI Tools in Teaching and Assessment Working Party and the Centre for Academic English, adapted from Russell Group guidelines — are framed around promoting critical use, a consistent ethical approach, and building a research community. They contain no detection requirement and no declaration requirement.

So the “authorisation” in clause 14(v)(d) is departmental authorisation. Find your department’s policy, and where it is silent, ask and keep the reply. On what a good acknowledgement contains, our AI disclosure statement guide covers the shape — but follow your department’s form if it has one.

Before you submit

• Find your department’s AI policy. Imperial publishes none centrally and tells you to look there first.
• Treat translation tools as in scope. They are named in both definitions.
• Acknowledge and get authorisation. The wording of the two limbs differs, so do both.
• If you are a research student, read Ordinance D17 — this policy does not apply to you.
• Keep drafts and version history. An allegation can come at any point, including after an award.

If you want a read on your own text before you hand it in, our AI check is free up to 1,500 words, with no account and no name.

Sources

• Imperial College London: Academic Misconduct Policy and Procedure — v1.3, effective 15 March 2023; first approved 20 June 2018 by Senate.
• Imperial College London: Examination and Assessments: Academic Integrity — Version 4, September 2018, approved by Senate; not revised since.
• Imperial College London: Generative AI Principles — last updated 5 March 2025.

This page is orientation, not legal advice, and it covers taught programmes. What binds you is your department’s policy and the Academic Misconduct Policy — or Ordinance D17 if you are a research student.