The governing principle
Use AI to improve your ability to perform research—not to conceal that you did not perform it. Verify and defend every material claim, source, calculation and decision.
Lower- and higher-risk uses
Generally lower-risk
- Generating questions to test an argument.
- Developing search-term variants that are independently tested.
- Obtaining an introductory explanation before consulting authoritative sources.
- Checking clarity while preserving the researcher’s reasoning and voice.
Higher-risk
- Submitting generated literature, analysis or interpretation without verification.
- Using invented or unverified citations.
- Uploading confidential or identifiable data without authority.
- Delegating decisions the researcher is expected to defend.
- Using AI where institutional instructions prohibit it.
A verification workflow
Confirm permission
Identify what the institution, supervisor, assessment and intended journal permit.
Protect information
Remove personal, confidential, unpublished and proprietary data unless authorised safeguards exist.
Treat output as unverified
Locate original sources and check every material factual or methodological statement.
Reconstruct the reasoning
Ensure the researcher understands, edits and can defend the final decision.
Disclose where required
Record the tool, purpose and extent of use in the form required by the institution or publisher.
Institutional policy controls
Policies differ across universities, programmes, supervisors and journals. Confirm the applicable rule before using AI on assessed or publishable work. When rules conflict, apply the more specific instruction governing the submission.