Disclosing AI

It all started with a post we read on social media.Reading online criticism of authors and bloggers whose pieces had been flagged by AI-screening tools as entirely machine-written, we found ourselves agreeing with the impulse behind the criticism while questioning its target.We use AI extensively in our daily practice, yet we do not feel the guilt these critiques seem designed to induce because, we thought, 'the work we do with AI is not "write this for me"'.It is an active process.We set out what must be said, react to a draft, redirect emphasis, correct substance, and repeat that cycle until the text reflects an argument we already held before the exchange began.The finished prose and a single-prompt output might look identical to a detection tool, but they represent entirely different relationships between author and machine.That gap made us reflect deeply.Generative AI has become embedded in the daily workflow of scientific writing, from language editing to full manuscript drafting.Journal policy has responded with a single, blunt instrument: disclose whether AI was used, and for what general purpose.The International Committee of Medical Journal Editors' (ICMJE) January 2026 update formalized this into a dedicated section of its Recommendations, requiring authors to disclose at submission whether AI-assisted technologies were used and to describe how, in both the cover letter and the manuscript itself [1].The Committee on Publication Ethics (COPE) and most major publishers converge on the same principle.What none of these frameworks provide is a way to distinguish how AI was used, and this omission can, in our opinion, substantially affect how AI-assisted authorship is perceived and evaluated.