Generative AI
Language and other models generate text, images and code based on patterns learned from training data. They do not “understand” facts and can produce false information with high confidence (hallucination).
AI in literature discovery
AI search tools can suggest and summarise sources, but always go back to the original source, read it, and check that it exists and says what is claimed.
AI for research support
- Language editing and proofreading.
- Suggesting an initial structure or review questions.
- Helping to write analysis code, which you must test carefully.
Responsible AI
Know the tool’s limits and biases, and take responsibility for everything you publish. A tool is not an author and should not be listed as one, in line with major publishers’ positions.
Research integrity
Disclose tool use as the journal requires (in methods or acknowledgements). Never fabricate data, images or references.
AI and academic writing
Use it to improve language, not to generate ideas and results. Rewrite suggestions in your own voice and make sure no text is copied.
Verifying AI-generated references
- Look up each reference in a trusted database or via its DOI.
- Compare title, authors, year and journal.
- Actually read the source before citing it.
- Remove any reference you cannot verify.
Privacy and confidentiality
Do not enter personal data, manuscripts under review or confidential text into public tools; they may be stored or used to train models. Check the tool’s policy and the policy of the body that asked you to review.