How to Disclose AI Use in an Assignment
More courses now allow AI and require you to say how you used it. What a disclosure statement is, what to put in it, a template you can adapt, and the two mistakes that cause trouble.
For two years the question students asked was how to avoid being accused. That question is being replaced by a different one. A growing number of courses now permit generative AI in coursework on one condition: that you say how you used it. If your assignment brief mentions acknowledging, declaring, or disclosing AI, it is asking for a short statement, and there is a right way to write it.
This guide covers what that statement is, what belongs in it, and the two mistakes that turn a routine formality into an academic integrity meeting.
Why disclosure replaced detection
Institutions moved because detection stopped being defensible. Australia's national higher education regulator now advises universities that reliably detecting AI in unsupervised work is not achievable, and directs them to redesign assessment rather than spend on detection tools. Its guidance puts the process of learning at the centre, on the reasoning that process is the part AI is least able to simulate.
What replaced blanket bans is a split. Some assessments are secured in person, supervised, or spoken aloud, where the question of AI does not arise. The rest are left open, and in the open ones the rule is not that you must avoid AI. It is that you must be straight about it. The University of Sydney, whose two lane model has been widely copied, went as far as removing the option for course coordinators to prohibit AI in open assessment, while requiring students to acknowledge how they used it.
So the skill worth having is no longer concealment. It is describing your own process accurately.
What a disclosure statement actually is
It is short. Three or four sentences is normal, and anything longer usually means you are justifying rather than describing. It goes wherever your brief says, and if the brief does not say, a note at the end of the document above your references is the safe default.
It has two jobs. The first is to name what you used AI for, specifically enough that a marker can picture it. The second is to make clear what remains yours: the argument, the sources, the judgement about what matters. A statement that does both reads as a competent researcher describing a workflow, which is exactly what it should read as.
What to include
Name the tasks, not the tools. Marking staff care that you used AI to plan a structure or to check grammar, far more than which product you opened. Tasks are also stable: the tool you used may be gone next year, but planning and proofreading mean the same thing to any reader.
Be specific about scope. There is a real difference between using AI to check the grammar of paragraphs you wrote and using it to draft passages you then revised. Both are commonly permitted. Only one of them is what a vague sentence like I used AI for help implies, and the vagueness invites the marker to imagine the larger version.
Say what stayed yours. One clause is enough: that the argument, the reading, and the choice of sources are your own. If any of that is not true, the disclosure is not the problem you need to solve.
A template you can adapt
Use this as a shape, not a script. Marking staff read a great many of these, and identical wording across a cohort is noticeable.
AI use disclosure. I used generative AI to plan the structure of this essay and to check grammar and clarity in passages I had already written. All of the argument, the sources, and the analysis are my own. I did not use AI to draft the text.
If you did use it to draft, say that instead, and say what you did with the output. Drafting a passage and then rewriting it in your own words is a normal, disclosable workflow in most open assessments. Presenting it untouched usually is not.
The two mistakes that cause trouble
The first is under-declaring. A statement that says you did not use AI, submitted with work that shows evidence of it, is worse than any amount of disclosed use, because it converts a permitted act into a dishonest one. This is the mistake that turns an ordinary submission into a misconduct case.
The second is over-claiming in the other direction: statements that promise the work is one hundred per cent human, or that no AI was involved at any point, when the writer has used a grammar checker, a translation tool, or an in-editor suggestion. Most of those are AI. A precise sentence about what you used beats a sweeping denial that you cannot support.
Between the two, the safe habit is the same: describe what happened, at the level of detail you would be comfortable being asked about.
Evidence makes a disclosure believable
A statement is a claim. What makes a claim easy to accept is that it is specific and that it is consistent with everything else the marker can see. If you wrote the piece across several sittings, revised as you went, and typed most of it yourself, saying so, with figures, is far stronger than saying it in the abstract.
This is also why keeping some record of your process is worth the small effort. Drafts with version history, notes, and a writing record all serve the same purpose: they let you describe your own work concretely rather than from memory, months later, under pressure.
Check your own brief first
Policies differ by institution, by faculty, and sometimes by individual unit, and they are changing quickly. Everything above is the shape of a good disclosure, not a substitute for the rule that applies to you. If your brief specifies a form, a location, or a required wording, follow it exactly. If it says nothing at all, a short accurate note costs you nothing and answers the question before it is asked.