Literature

• AI-assisted screening in a systematic review and where PRISMA still binds
• Verifying every retrieved reference against the primary source
• The narrowing effect of letting a model choose what you read

Analysis

• Generated analysis code that runs but is silently wrong
• Confidentiality: what must never be pasted into a third-party service
• Reproducibility when part of the workflow is a non-deterministic model

Writing

• ICMJE and COPE positions: AI cannot be an author and use must be disclosed
• Publisher and funder policies and how sharply they differ
• AI-detection tools and their false-positive rates against non-native writers

Governance

• Institutional policy, research integrity and what counts as misconduct
• Attribution, plagiarism and derivative text in generated output
• A personal use policy recorded so decisions can be defended later