Can AI Be an Author? The Rules for Using AI in Research and Publishing

When manuscripts began listing chatbots as co-authors, editors converged quickly on the same answer: AI tools cannot be authors, because authorship requires accountability that software cannot bear. The humans who use these tools stay fully responsible for the work, and current guidance asks them to disclose which tool they used and how.

The question that forced a policy

When generative language tools became widely available, some early manuscripts listed a chatbot as a co-author. Editors had to answer a question they had never faced: can a piece of software be an author? Within a short span the major bodies, including the International Committee of Medical Journal Editors and the Committee on Publication Ethics, converged on the same answer. No.

The agreement was notable because it was quick and broad. It did not rest on fear of the technology. It rested on what authorship means.

Why accountability rules out AI authorship

Authorship carries responsibility. An author must be able to stand behind the accuracy and integrity of the work, to declare conflicts of interest, to approve the final version, and to answer if the work is questioned. A software tool cannot do any of these. It cannot be accountable, cannot hold or assign copyright, and cannot be called to explain itself in any meaningful sense.

So the reasoning is not that the tool contributed too little. It is that authorship is a form of accountability the tool is incapable of bearing. The humans who used it remain fully responsible for every part of the manuscript, including anything the tool produced.

Disclosure: what authors are asked to say

Ruling out AI authorship does not mean pretending the tools were not used. The guidance is the opposite: be transparent. Authors who use these tools are asked to disclose it, naming the tool and describing how it was used, typically in the methods or acknowledgments depending on the task.

The logic mirrors any other tool or assistance. If a language model helped draft text, that belongs in the acknowledgments alongside other writing help. If it was used to analyze data, that belongs in the methods, where a reader can weigh it. Transparency lets others judge the contribution rather than be misled by its absence.

AI for reviewers and editors

The policies extend past authors. Peer reviewers are generally told not to upload a manuscript they are reviewing into a public tool, because doing so can breach the confidentiality of unpublished work and the authors' rights over it. Editors carry responsibility for ensuring the tools are used responsibly across the process.

This part is easy to overlook and important. A confidential manuscript pasted into an external service may leave the author's control. The guidance treats the peer review file as privileged material, whatever tool is in front of the reviewer.

Where the tools can and cannot help

None of this is a blanket ban on using the technology. Used openly, language tools can help with phrasing, translation, and readability, which can genuinely lower barriers for researchers writing in a second language. The line is drawn at accountability and disclosure, not at use.

What the tools cannot do is supply the judgment authorship represents. They can generate fluent text that is confidently wrong, invent citations that do not exist, and smooth over gaps a human expert would catch. The requirement that a named person verify and take responsibility for the content is exactly the safeguard against those failure modes.

Reading a paper in an era of generated text

For a reader, the practical shift is modest but real. Look for a statement disclosing whether and how AI tools were used. Its presence is a sign of a journal and authors following current norms; its absence, in a venue that requires it, is worth noting.

The deeper point is unchanged. Fluent writing was never evidence of a sound study, and it is even less so now. Judge a paper by its methods, its data, and whether named, accountable people stand behind it, which is the same standard that predates any of these tools.