You used an AI feature to suggest a structure, explain a difficult passage or comment on your wording. Now the submission asks about assistance, and “I used AI” feels too vague. AI disclosure for university assignments should explain what happened clearly enough for someone to understand the tool’s actual contribution.
This guide offers original examples and a practical record-keeping method. It does not authorise AI use. Check the instructions for your specific assessment first: permission in one subject, institution or task does not automatically apply to another.
What should an AI disclosure include?
An AI disclosure should identify the tool, explain its permitted purpose, describe the work it affected and state what you did with its output. Include dates, prompts, versions or evidence where your assessment requires them. Follow the institution’s own format, and never use an acknowledgement to disguise an unauthorised use.
Guidance checked on 5 September 2026: institutional requirements can change. The sources linked below are examples of university guidance, not a universal policy or an endorsement of a particular tool.
Check permission before writing a declaration
Start with the assessment brief and current subject instructions. Identify the specific activity: brainstorming, translating, rewriting, generating code, checking grammar or producing an image. A rule about one activity may not cover the others.
The University of Melbourne instructs students to check authorised uses with their subject coordinator. Its guidance also distinguishes acknowledging assistance during the process from citing AI output that is incorporated into the work. Follow your own institution’s requirements rather than importing Melbourne’s rules into another course.
If the brief is unclear, ask a narrow question before using the feature: “May I use an AI tool to identify grammatical errors in my own draft, without generating replacement paragraphs, and how should I record that assistance?” Save the answer with your assignment materials. A generic “AI is fine” conversation may not resolve the boundaries of the actual task.
Acknowledgement and citation do different jobs
An acknowledgement describes assistance with the process. A citation identifies material used as a source according to the required referencing style. Depending on the task and the use, you may need one, both or a different institutional declaration.
Monash University’s guidance on acknowledging AI asks students to follow assessment restrictions and explain the tool’s contribution and extent. It notes that the required evidence can include more than prose, such as screenshots or recordings. The exact task instructions therefore matter as much as the wording of the final statement.
Do not add a tool to the reference list and assume that this explains how it shaped your work. Equally, an acknowledgement does not replace citations to the research sources supporting your argument. Read those sources yourself and check that they genuinely support the claims being made.
Keep a small record while you work
Here is our suggested working log. It is a practical aid, not an official declaration form. Store it privately with the draft history and adapt it to the evidence your course requests.
- Tool and feature: the product and the function you actually used; record a model or version only if known.
- When: the date of the interaction or working session.
- Purpose: the specific problem you asked it to address.
- Input: the relevant prompt or instruction, retained safely where appropriate.
- Output: what it suggested or generated.
- Your decision: what you rejected, checked, changed or included.
- Location: the section, figure, code file or stage affected.
For example, “planning help” leaves too much unspecified. A more useful log says that the tool suggested three ways to group an argument, you rejected two because they missed the question, and you developed the remaining structure using your own reading. Record the real sequence, not the tidiest story you can reconstruct afterwards.
Three disclosure examples to adapt honestly
These are fictional examples, not pre-approved wording. Use them to understand the level of detail. Do not copy an activity, claim or verification step that you did not perform. Replace bracketed fields only with accurate information, and use your university’s template where one exists.
Example 1: permitted brainstorming
During planning on [date], I used [tool and feature] to suggest possible subtopics for [task]. I used the suggestions as questions to investigate, not as evidence. I selected the final scope after reading the sources cited in the assignment and wrote the submitted analysis myself. [Describe any other assistance, if applicable.]
This example distinguishes idea generation from research evidence. It would be misleading if the tool also drafted paragraphs that appear in the submission.
Example 2: permitted language feedback
After writing my draft, I used [tool and feature] on [date] to identify unclear wording in [sections]. It suggested [describe the actual changes]. I accepted [specific types of changes] and rejected [specific suggestions]. The retained changes affect [locations]. [State any additional drafting or translation assistance accurately.]
A statement about spelling alone is inaccurate if the feature restructured the argument or rewrote sentences. Describe what happened rather than relying on a product’s broad “writing assistant” label.
Example 3: permitted code assistance
I used [tool and feature] on [date] to help investigate [specific issue] in [file or function]. Its output suggested [approach]. I incorporated [precise contribution], changed [details] and checked the result using [tests actually run]. The relevant interaction record is provided in [required location].
Testing whether code runs is different from checking whether it answers the assignment correctly. Record actual checks without upgrading a successful execution into proof of correctness.
Handle awkward cases without minimising them
If you used an embedded AI feature without noticing at first, record what you now know: the application, feature, affected material and available history. Ask the teaching team how to proceed. Do not invent prompts or precise model details that the application never displayed.
If your use went beyond the permitted scope, a polished declaration does not make it acceptable retrospectively. Stop expanding that use, preserve the relevant records and seek advice from the course team or student support service. Whether revision or another process is appropriate depends on your institution.
If no AI was used, follow any required no-use declaration. Make sure the statement is accurate about both generation and other assistance; do not deny all use simply because you did not paste generated paragraphs into the final document.
Protect information and verify what remains
Before uploading material, check institutional restrictions and the tool’s data-handling arrangements. Avoid sharing identifiable participant data, classmates’ work, confidential placement information or restricted teaching materials without appropriate permission. Keep evidence through approved storage rather than making a chat publicly accessible by default.
Before submission, compare the disclosure with your draft history. Does it cover the actual features and affected sections? Have you checked the retained references, calculations and factual claims? Our research-source guide explains how to verify an article instead of trusting a plausible-looking citation.
Frequently asked questions
Will declaring AI use guarantee that it is accepted?
No. Transparency and permission are separate requirements. A declaration explains use; it does not override assessment restrictions or guarantee a grade.
Do I need to include every prompt?
Follow the assessment instructions. Some tasks request a concise summary; others require a fuller process record. Keep relevant evidence while working so you can provide what is required accurately.
Is a similarity score an AI-use declaration?
No. A text-matching report and a description of assistance serve different purposes. Do not infer permission or compliance from a low score. Read our similarity-score guide for the separate question of interpreting source matches.
Need help understanding the next step?
Ask about tutoring or feedback on your own draft. Tell us what your course permits; you remain responsible for the research, reasoning and submitted work.
Discuss study support