Plagiarism and AI at Sheffield: AI-free assessments and what counts

The short answer

At Sheffield the rule is set per assessment, and some assessments are labelled AI-Free — meaning no GenAI use at all. Claiming authorship of AI-generated content is a named offence called automated authorship.

The governing document is the Academic misconduct policy — taught students, approved by Senate on 11 December 2024. Note the terminology shift: Sheffield has replaced “unfair means” with “academic misconduct”, though some older page addresses still carry the old term.

AI-Free assessments

This is the practical thing to check, and it is stated in one line:

“AI-Free, in which case GenAI tools must not be used to support your work.”

Note “to support”. That is broader than “to write”. In an AI-Free assessment, using a tool to brainstorm, outline or tidy your prose is outside the permission, not just using one to draft the argument.

Where use is permitted, disclosure is required — Sheffield asks you to “fully disclose any AI-generated content used” and describes “full acknowledgment of GenAI involvement required”. There is one wrinkle worth knowing: students are told not to cite GenAI as an author or as a source. So the acknowledgement is a statement about your process, not a bibliography entry. Our AI disclosure statement guide covers that shape.

Sheffield does not publish a prescribed form of words, so keep it simple and specific: which tool, for what, on which parts.

Automated authorship

Sheffield gives AI its own category rather than folding it into plagiarism:

“Automated authorship, involving the inappropriate/unsanctioned use of Artificial Intelligence.”

And on the student-facing page, more plainly:

“Claiming authorship of AI generated ideas or content is considered false (automated) authorship and a form of academic misconduct.”

The operative word in the formal definition is unsanctioned. Sheffield is not saying AI is forbidden; it is saying the sanctioned scope is defined for your assessment and anything beyond it is misconduct. Which brings you back to checking the classification.

Sheffield does not use detection tools

If you are ever accused on the strength of a percentage at Sheffield, this is the sentence to have ready. It is published in two places, and the fuller version reads:

“GenAI detection tools are not used at the University of Sheffield. This is due to concerns over their error rates and the potential for both false positives and false negatives.”

It names the reason precisely: errors in both directions. A tool that misses AI writing and a tool that accuses human writers are two different failures, and Sheffield says it is concerned about both.

One honest limit: neither page carries a date, a version or an owning office, so we cannot tell you when this was decided or by whom. Sheffield does have institutional “Principles for generative AI in education”, but they sit behind a university login and we could not read them. If you need to rely on the position formally, ask your department to confirm it in writing.

Sheffield does use Turnitin for similarity — “We use a range of strategies to detect academic misconduct, including Turnitin, which helps detect plagiarism” — on a page that is also undated. What a similarity percentage does and does not tell anyone is covered in what a Turnitin score actually means, and the wider reliability question in how accurate AI detectors really are.

And it tells you not to use one either

We should quote this even though it is inconvenient for us, because you need to know it:

“Do not use non-University-approved software to check your work for plagiarism before submission.”

That covers third-party checkers generally — ours included. If you study at Sheffield, take the instruction at face value and ask your department what the approved route is. The wording implies one exists; the page describing it is behind a login, so we cannot tell you what it is.

The reasoning is not stated, but the usual concern is real and worth knowing wherever you study: some checking services store what you upload and compare future submissions against it, so your own draft can come back as a match against you. Before you put work into any service, find out whether it keeps it. For what it is worth, our own check does not add your text to any third-party comparison database — but at Sheffield, the university’s instruction is the one that governs.

The other categories

Alongside automated authorship, Sheffield names:

• Plagiarism— “using the ideas or work of another person and submitting them as the student’s own”.
• Collusion — “working with others to produce a piece of work, all or part of which is then submitted by each student as their own individual work without acknowledgement”.
• Contract cheating — “submitting bought or commissioned work from a third party, including essay mills, other students, friends or family members”.
• Self-plagiarism, fabrication, impersonation and examination cheating.

That contract-cheating definition catches something people rarely think of as cheating: a friend or a relative writing part of it for you. No money need change hands.

Sheffield’s own definition of misconduct is broad — “any act that compromises the integrity of the academic process, and/or has the potential to gain an unfair advantage” — and note “has the potential to”. Actual advantage is not required.

Three severity bands

• Poor academic practice — an educational response, no penalty. This is where genuine referencing confusion is meant to land.
• Moderate offences — formal warnings; the component or module awarded 0%; possibly a resubmission or resit with the mark capped.
• Severe offences — 0% with no resit opportunity, and expulsion, which only a Senate Discipline Panel may impose.

Two procedural protections are published. An investigation must not exceed 30 working days from notification. And a penalty cannot be applied without Faculty Officer approval — your marker cannot simply impose one.

Appeals are handled by the Student Conduct and Appeals Team. If you are facing a case, what to do when you are accused of using AI and how to prove you wrote it yourself 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.

What Sheffield does not publish

We would rather tell you what is missing than guess. On the public web we could not establish:

• Whether submissions are added to a comparison repository, or any opt-out.
• Whether a university-approved student self-check route exists.
• The standard of proof applied.
• Whether a Completion of Procedures letter and the Office of the Independent Adjudicator stage apply — they normally would at a UK university, but Sheffield does not say so on these pages.
• Any power to revoke a degree.
• Any annual case counts.

Several of Sheffield’s student pages sit behind its login, so some of this is likely documented where you can see it and we cannot. If a point matters to you, it is worth signing in and looking.

Before you submit

• Check the AI classification for that assessment. AI-Free means no GenAI support at all.
• Disclose any permitted use — but as a process statement, not as a citation. Sheffield asks you not to cite GenAI as a source.
• Do not use an outside plagiarism checker. Sheffield says so directly.
• Do not let a friend or family member write any of it. That is contract cheating in Sheffield’s definition.
• Keep drafts and version history. With no detector in play, a question about your work will be answered by your work.

Sources

• University of Sheffield: Academic misconduct policy — taught students — approved by Senate, 11 December 2024.
• University of Sheffield: AI and academic integrity — the statement that GenAI detection tools are not used; page carries no date.
• University of Sheffield: Unfair means — Turnitin, AI-Free assessments, and the advice on outside checkers; page carries no date.

This page is orientation, not legal advice. What binds you is the assessment brief and Sheffield’s academic misconduct policy.