How to Use AI Responsibly in Medical Research

Use AI tools in medical research as assistants whose output you must verify, never as sources of truth or as authors. Follow the rules of your institution, funder and target journal, protect confidential and patient data, disclose how you used the tools and take full responsibility for every statement you publish.

Principles

1. You are accountable. Authors, not tools, are responsible for the accuracy, originality and integrity of the work.
2. Check journal and institutional policy on AI use before you start; policies differ and change.
3. Protect confidentiality. Do not enter patient data, unpublished data, confidential manuscripts or peer-review material into tools that are not approved for that purpose.
4. Verify everything. AI can generate fluent but false statements and fake references.
5. Disclose use as the journal requires, usually in the methods or acknowledgements, naming the tool and how it was used.
6. Do not list AI as an author. (See Can AI Be Listed as an Author on a Research Paper?)

Lower-risk uses (with verification)

• Brainstorming search terms.
• Language editing of your own text.
• Summarising a paper you have read, as a check against your understanding.
• Drafting code, then testing it.

Higher-risk uses

• Generating references or citations.
• Writing methods or results you did not carry out.
• Analysing data without understanding the method.
• Interpreting clinical data for decisions.

Illustrative example

Invented for teaching. A student asks an AI tool to suggest synonyms for a PubMed search, then tests each in PubMed and keeps those that retrieve relevant papers. She notes in her methods: "An AI tool was used to suggest search terms; all terms were verified by the authors."

Common pitfalls

• Pasting confidential data into a public tool.
• Trusting a plausible answer.
• Failing to disclose.

Key takeaways

• Verify, protect data, disclose and remain accountable.
• AI is an assistant, not an author.

This page is general educational information about research methods. It is not medical, legal or regulatory advice; follow your institution's ethics board and your target journal's instructions.