I keep running into: not whether we use AI, but whether we’re honest about it.
I stumbled upon a paper the other day, Provenance, Not Prohibition, and the title alone made me pause. Most journals are still stuck arguing about bans, as if we could detect AI writing reliably enough to enforce one. We can’t. The line between “AI helped me phrase this” and “AI wrote this” was never going to hold, and chasing it wastes energy that should go somewhere more useful.
The paper’s real argument is simpler than the title suggests: document what the AI did, and keep the human accountable for what goes out under their name. Literature search, drafting, code, editing, whatever it was, write it down. That’s provenance, and it’s a far more honest ask than pretending the tool wasn’t in the room.
Here’s what I’d add, having spent enough years around research to know where the gaps hide. A line in the methods section saying “AI assisted with this manuscript” tells a reader almost nothing by itself. It doesn’t say whether the citations were checked, whether someone reran the analysis by hand, whether the draft was actually read before it went out. Disclosure moves the trust question one step back, from “was AI used” to “do I trust this author’s word that they checked it.” That’s progress. It isn’t the whole answer.
If I were writing policy for a journal, I’d grade disclosure by role rather than treat it as one box to tick. AI helping with a literature scan carries a different risk than AI touching the statistics, and most current guidance, COPE and ICMJE included, doesn’t yet make that distinction. Provenance is the right instinct. It needs sharper teeth.
Source: Johnson RF. Provenance, Not Prohibition: A Framework for AI in Scholarly Publishing. Laryngoscope Investigative Otolaryngology. 2025. onlinelibrary.wiley.com/doi/10.1002/lio2.70487