Responsible AI Use in Research Writing: Assistance, Accountability and Disclosure
Use AI tools without surrendering authorship, verification or responsibility.
The author remains accountable
AI systems cannot take responsibility for accuracy, conflicts, consent or integrity and therefore should not be listed as authors. Human authors must verify generated statements, references, calculations and code.
Appropriate use depends on journal, institution and funder policy. Language improvement may be allowed while undisclosed generation of analysis, evidence or peer-review content may not be.
Protect confidential material
Do not upload unpublished manuscripts, personal data, proprietary results or confidential peer-review documents to a service unless its terms and your authorization permit that processing. Remove sensitive identifiers where possible.
Keep a record of the tool, purpose, relevant prompts or transformations and human verification performed. Disclose use when required and when it materially affected the work.
Verify everything
Check every cited source, numerical claim and quotation against a reliable original. AI-generated references can be plausible but nonexistent. Re-run code and calculations in the intended environment.
Use AI as assistance, not evidence. The scholarly contribution, reasoning and final decisions must remain genuinely attributable to the authors.