Not exactly — and the distinction matters more than it sounds. Plagiarism means presenting someone else's work as your own. AI-generated text has no author to steal from, so most UK universities now treat submitting it as its own category of misconduct: unauthorised assistance, unfair means or contract cheating, depending on the wording.
The outcome is usually the same. But it means “I didn't copy anyone” is not a defence, because copying was never the allegation.
Why universities rewrote the rule
Traditional plagiarism regulations are built around a source: a paper, a book, another student's essay. They ask whether you reproduced someone's words without credit. Run an AI-written essay through that test and it passes — a plagiarism checker will often return a low similarity score, because the text is genuinely new.
That gap is why most UK institutions added separate wording between 2023 and 2026. The thing being assessed is your understanding, and submitting text you didn't write misrepresents it regardless of whose words they technically were.
A low similarity score is not evidence you're in the clear. Similarity and AI detection are two different reports measuring two different things — see does Turnitin detect ChatGPT? for what each one actually does.
Where it does become plagiarism
Two routes, and both are more common than students expect:
- Regurgitated training data. Models sometimes reproduce phrasing close to their sources. If that lands in your essay uncited, it's ordinary plagiarism and a similarity checker will find it.
- Fabricated references. A model will produce a plausible citation — real author, real journal, article that was never written. Submitting it claims you read something you didn't, which most regulations treat more seriously than the AI use itself. More on that in can I trust ChatGPT for references?
What actually crosses the line
Roughly, and subject to your own course's wording:
- Usually fine: having a concept explained, generating angles to consider, checking grammar and spelling, sanity-checking a structure — then writing it yourself.
- Misconduct almost everywhere: submitting generated paragraphs as your own words; running AI output through a paraphraser to disguise it; citing sources you never read.
- Depends, so ask: AI-summarised sources you then cite, translation, and heavy redrafting of a generated first draft. A growing number of courses permit these if declared — and treat undeclared use as the breach.
How serious is it?
It scales with what you did and whether it's a first instance. Typical outcomes run from no case to answer, through a capped or zeroed mark on the assessment, up to failing the module and — for repeated or serious cases — exclusion. Nearly every institution has an internal appeal, and once that's exhausted the Office of the Independent Adjudicator can review the process.
If you're already facing an allegation, this is what to do first.
SafeGrade checks your essay for AI-looking passages and referencing problems, and Sage can tell you what your specific course permits. Private, never stored.
Run one free scan →No card. No catch. First scan free.