Reading your actual policy

The rule that applies to you is written down somewhere

Students ask is it allowed and answer it from what a friend said, what a lecturer implied, or what a video claimed. The rule that will be applied to you is a document, and finding it takes ten minutes once.

Look, in this order:

1. The module handbook. Course-level rules override general ones and are frequently stricter or more specific.
2. The assessment brief for this piece of work. The most specific rule wins, and briefs sometimes carry a line that contradicts the general policy.
3. The academic integrity or academic misconduct policy, usually on the university’s registry or quality-assurance pages.
4. Any separate generative-AI guidance, which many institutions issued from about 2023 onwards and revise often.

Download them. Policies get updated, and the version that applied when you submitted is the one you want to be able to produce.

The three-tier structure most policies use

Despite enormous variation in wording, most institutions have converged on some version of three categories, often called a traffic-light or scale approach:

• Not permitted. No AI at any stage. Common for language assessments, some exams, and any task where the thing being tested is exactly what a model does.
• Permitted with disclosure. Use it, say how. The largest category and growing.
• Permitted, encouraged, or required. Increasingly common in professional and technical subjects where using the tools is part of the competence being assessed.

A well-written brief tells you which tier applies. If it does not, that is the question to ask.

Words that matter, and what they usually mean

Policies are written carefully and read carelessly. The distinctions that decide cases:

• ”Generate” versus “assist”. Producing new text for your submission versus improving text you wrote. Almost every policy treats these differently, and it is the main line.
• ”Substantial” or “significant”. Usually undefined. Ask what it means for your assessment; the answer in writing is worth having.
• ”Proofreading” and “language support”. Usually permitted, sometimes with a disclosure requirement, and sometimes limited to correcting errors rather than rewriting.
• ”Your own work”. The oldest phrase in the document, and the one everything else is interpreting.
• ”Undisclosed”. Several policies make non-disclosure the offence rather than the use. Under those, a permitted use becomes misconduct by not being declared, which is a trap for the honest.

When the policy is silent or unclear

This is common, and the answer is the same every time: ask in writing, and keep the reply.

Dear Dr X, for the coursework due on the 14th, could you confirm whether it is acceptable to use an AI assistant to (a) generate practice questions while revising, (b) check my draft against the mark scheme, and (c) correct grammar in my own prose? I want to make sure my disclosure statement is accurate.

Three things this achieves. You get an answer. You have it in writing, which is evidence. And the framing signals that you intend to disclose, which changes how the question is read.

Email, not a corridor conversation. A remembered conversation is not evidence and honest people misremember.

The safe default when you cannot find out

If you cannot establish the rule before a deadline:

• Use it only for things that produce no text in your submission: practice questions, explanation, structural diagnosis of your own draft.
• Write every word yourself.
• Disclose what you did, briefly and factually.

That combination is inside almost every policy in existence. It is not the most efficient possible use; it is the one that cannot go wrong.

Policies vary more than you would expect

Between countries, between institutions, and between departments in one institution. A computing department may require AI use; the law school next door may prohibit it entirely; the language centre may prohibit it for one assessment and encourage it for another. All of that can be true simultaneously and consistently, because they are assessing different things.

So do not import a rule from another course, another university, or the internet. The document that applies to you is the one for your module, and the fact that a friend on another programme does something different is not evidence about your position.

And the honest part

These policies are new, they are being revised constantly, and many of them are internally inconsistent or unenforceable as written. Institutions are working it out in public. That is a real situation and not a reason to ignore the rules; it is a reason to document what you did, ask when unsure, and disclose generously. Those three behaviours protect you under any version of a policy, including the one your institution has not written yet.