Before using AI in an assignment, read the rule for that assignment

Permission can differ between practice, editing, coding and assessed work. An honest process record helps students explain permitted assistance.

Your university may provide an AI account while a particular assignment restricts its use. Those statements can both be true. Access gives you a tool; the assessment rule tells you what work you must do and what assistance is permitted. Read that rule before using AI, because repairing an unclear process after submission is harder than making a clear decision at the start.

Oxford’s [student AI questions](https://www.ox.ac.uk/students/life/it/genai-tools/genai-faqs-for-students) discuss responsible use and assessment expectations. Cambridge’s [assessment guidance](https://blendedlearning.cam.ac.uk/artificial-intelligence-and-education/generative-ai-and-assessment) supports discipline-specific decisions. Oxford’s [research policy](https://www.ox.ac.uk/research/support/governance-and-committees/research-policies/policy-for-using-generative-ai-in) separately addresses research responsibilities and declarations. These are institutional examples checked in October 2026. They are not a universal permission for students elsewhere or a substitute for a current module instruction.

Start by identifying the task. Is it private practice, formative feedback, a marked essay, code, a presentation or a research submission? Then find the written guidance for that course and assignment. If it does not answer the proposed use, ask the teacher or responsible department before relying on it. Do not treat a friend’s interpretation or a chatbot’s answer as authority.

An original illustrative essay permits feedback on a student’s outline but forbids generated submitted prose. You write the outline yourself and ask an approved assistant to identify unclear connections. You read the comments, decide which are useful and make your own revisions. Asking it to write the final essay would be a different use, even if you later changed some words.

A coding assignment may draw the boundary elsewhere. The course could permit syntax help while assessing your algorithm design. In that setting, describe the error and ask for an explanation rather than requesting a completed solution. Test any suggested change and explain why it works. If the course prohibits that assistance too, use the authorised support channel instead.

A short process record can prevent confusion. Note the tool, date, task, type of assistance and the changes you chose to make. Keep relevant drafts or permitted interaction excerpts when the course asks for them. Avoid collecting sensitive material unnecessarily. The record should describe the actual use, not minimise it with a vague statement that AI only helped.

Disclosure does not create permission. If an activity was prohibited, acknowledging it afterwards does not automatically make it acceptable. Conversely, a permitted activity may still require a declaration in a particular format. Follow the course’s instructions for acknowledgement and citation. A generic template from the internet may omit details the assessor needs.

You remain responsible for accuracy. AI can supply nonexistent references, incorrect code or a claim that the assignment instructions allow something they do not. Open the relevant sources, run permitted checks and preserve uncertainty where the evidence is incomplete. An answer saying this is academically acceptable is simply generated text unless it accurately refers to the actual rule.

Do not outsource the thing the assessment is intended to measure. That may be reading, argument, calculation, translation or original design. When a tool does that work, the resulting submission may cease to show your competence. A useful self-check is whether you can explain the reasoning, choices and evidence without the assistant.

Accessibility arrangements need the proper institutional route. A tool may help a student overcome a barrier, but the specific arrangement should be agreed with the relevant support service and course team. Do not assume that general AI restrictions settle every accommodation question, or that a disability automatically authorises every use. Clear communication protects both access and assessment integrity.

The benefit of permitted AI assistance can be better practice or feedback. The failure is submitting work that neither follows the rules nor represents your learning. Reading the exact instruction, asking early and keeping an honest process record reduces that risk. It also lets teachers distinguish authorised assistance from a misunderstanding without relying on a polished final file alone.

Before submission, review the work against the rule once more. Check the declaration, verify sources and identify any text or code you cannot explain. Resolve those gaps through the authorised course process. The aim is a submission you can defend as your work under the conditions your assessor actually set.