More than 90% written by an AI. If the exact level of detail of human input is not completely known, it will be sufficient if the paper is widely reported as being written by an AI. The journal has to be reputable (no pay to publish) and have a standard peer review process prior to publication.
Related questions
I am very interested in following AI, however, I believe that within a year they will not be able to develop it to the extent that no one will notice that it is not a human writing the article. Here should be the Critical Thinking Theory... questions and analyzes social structures, power, culture, and artificial intelligence. I came across an article https://papersowl.com/examples/critical-theory/ that discussed the importance of identifying and exposing inequalities and antagonisms in society with the aim of changing the status quo and creating a more just and egalitarian society. It's interesting to follow whether such cases will be revealed before 2025, depending on various factors, including the pace of artificial intelligence development, the rigor of expert review processes, and the transparency of scientific practices.
Possible Resolution?
"Performance of ChatGPT on USMLE: Potential for AI-Assisted Medical Education Using Large Language Models" is a paper about evaluating ChatGPT on the United States Medical Licensing Exam. When the preprint was published on MedRxiv, it got attention because ChatGPT got author credit. https://www.medrxiv.org/content/10.1101/2022.12.19.22283643v2
This was later discussed in a Nature report discussing whether AIs could be authors in research papers.
https://www.nature.com/articles/d41586-023-00107-z
The preprint was later accepted and published in the journal PLOS Digital Health.
https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000198
PLOS Digital Health is arguable a reputable journal: it is not pay-to-publish and has a peer review process. https://journals.plos.org/digitalhealth/s/editorial-and-peer-review-process
For what it's worth, "no pay to publish" and "standard peer review process" is a very low bar for academic journals. It includes many journals that I think would publish an AI-written paper even if it is poor quality. I'm reminded of the authors who got supposedly absurd papers published in their effort to critique postmodern and critical theory: https://en.wikipedia.org/wiki/Grievance_studies_affair#List_of_hoax_papers
@JacyAnthis True, but reproducing the grievance studies affair using an AI would not add much to human knowledge or make much of a point, IMO. It would be much more impressive to get an AI-written paper published in, say, Nature.