Disclosing AI use, by journal policy
Lesson 6 of 6 · Manuscripts and reviewer responses
1 · Learn the move · Flipped interaction
Journals genuinely differ on AI disclosure, and their policies move - what was acceptable at your last submission may be differently worded now. The stable frame comes from ICMJE and COPE: an AI tool cannot be an author, cannot take responsibility for the work, and its use is disclosed rather than hidden - where the disclosure lives (methods, acknowledgments, cover letter) is the venue's call. The flip: before drafting anything, the model interviews you, one question at a time, about every place AI touched the work - literature search, drafting, code, figures, statistics, editing - because the disclosure you write from memory is the one with the hole in it. The drafted disclosure then stays UNVERIFIED against your venue until you pull the journal's current policy page the week you submit. Policies change; that check does not transfer between submissions.
Interview me, one question at a time, about every stage where AI touched this manuscript: literature search, reference finding, drafting, editing, code, figures, statistics, translation. Then: (1) build the use inventory, (2) draft disclosure language in ICMJE/COPE framing - tool named, use described, responsibility ours, (3) mark the draft UNVERIFIED against [journal] until I confirm its current policy page, and list what to check there: placement, wording requirements, any prohibited uses.
2 · Your turn. You write the prompt
Submission is next week. Over ten months, a model helped debug the analysis code, drafted two paragraphs of the introduction that were heavily rewritten, and suggested references - some kept after verification. Three co-authors have three different recollections of what counts as disclosable, and the target journal updated its author guidelines sometime this spring.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.