What Should Students Disclose When Using AI Coding Tools?

A practical disclosure model

Students can disclose the tool or class of tool used, the purpose of the use, the material contribution, the important changes they made, and the verification performed. A short example might say: “I used a coding agent to propose a parser structure and generate initial tests. I rejected its error-handling approach, rewrote the interface, and verified the final behavior against these edge cases.”

This is more informative than attaching an unstructured transcript. It identifies where the tool mattered and what the student did with the output.

What counts as material?

Material use changes the solution, the reasoning, or the assessment evidence in a meaningful way. Examples include generated modules, substantial refactoring, agent-led debugging, generated tests that shaped the implementation, or a recommendation that affected an architectural decision. Trivial wording help or a quick syntax reminder may not need separate disclosure unless the course says otherwise.

When uncertain, students should be able to disclose briefly without being penalized for over-reporting. A simple “I used X for explanation and checked the result against the documentation” is often sufficient.

Transparency and privacy are both requirements

Students should not be expected to disclose private conversations, personal data, credentials, or confidential project material. Course guidance should say what must not be entered into an external tool and how submitted disclosure information will be handled.

Disclosure also should not become a hidden authorship test. It is one evidence source. The lecturer still assesses the artifact, reasoning, verification, and explanation against the learning outcomes.

Make disclosure easy to use

Use a short field in the submission form, a structured note in the portfolio, or a small section in the project report. Provide examples for permitted, prohibited, and ambiguous cases. If the form takes longer to complete than the learning activity it supports, students and lecturers will rationally treat it as bureaucracy.

Review the rule after the first course run. Ask whether the disclosures helped assessment decisions and whether they created unnecessary work. A good policy becomes more precise over time.

A sample disclosure form

Tool or tool category: ____. Purpose: ____. Material contribution: ____. What I changed or rejected: ____. Verification performed: ____. Remaining uncertainty: ____. This can be completed in a few minutes and gives the assessor a useful starting point without requiring a full transcript.

For a group project, add who used the tool and which shared component was affected. Keep the form focused on material work and allow students to say that no AI tool was used.

Explain the rule with examples

Clarity reduces both under-reporting and unnecessary paperwork.

Disclosure should not become a confession

Students may interpret a disclosure requirement as an admission that their work is less legitimate. Course teams can avoid that framing by explaining that professional developers document material tool use so colleagues can understand assumptions, review decisions, and reproduce important work. Disclosure is part of transparency, not an automatic mark reduction.

The same principle supports a student who used no AI. That student can state that no material AI tool was used and focus the evidence on the programming decisions the assignment assesses.

Review the policy for proportionality

Ask whether the required detail is necessary for the learning outcome and whether the lecturer can realistically review it. If the answer is no, reduce the form. A disclosure model that is too burdensome encourages vague or copied statements and gives assessors more text without more reliable evidence.

Proportionality is especially important for large introductory courses.

The disclosure conversation

Students should be able to ask whether a particular use is material before submitting. A course can provide a low-friction route: a short question field, an example library, or office hours. Treat honest uncertainty as a reason to clarify the policy, not as evidence of misconduct.

This makes disclosure a normal part of responsible development rather than a bureaucratic obstacle added at the end.