Updated 2026-09-23
from the news desk
Can AI be an author on a scientific paper? ICMJE and Nature say no
The International Committee of Medical Journal Editors says authors should not list AI as an author, and a Nature editorial says authorship brings accountability that AI tools cannot assume. Both also allow some disclosed AI assistance.
Under the International Committee of Medical Journal Editors recommendations and policies published by Nature and Springer Nature, AI should not be listed as an author on a scientific paper. Those documents allow some disclosed AI assistance, but they keep accountability, judgement, and responsibility with humans.
What the rules say
The International Committee of Medical Journal Editors, or ICMJE, says plainly that authors should not list or cite AI and AI-assisted technologies as an author. The same ICMJE section says journals should have a policy on AI use and that authors, reviewers, and editors should be transparent about when they use such tools.
Nature states the same bottom line for its own journals. In its 24 January 2023 editorial on tools such as ChatGPT, Nature says: “First, no LLM tool will be accepted as a credited author on a research paper.” Nature then gives the reason in the next sentence: “That is because any attribution of authorship carries with it accountability for the work, and AI tools cannot take such responsibility.”
Springer Nature's policy page also treats AI as a supporting tool rather than an author or decision-maker. The policy says scholarly judgement, accountability, and responsibility always remain human. Springer Nature summarizes that position in these words: “AI may support, but must not replace, scholarly judgement.”
Taken together, these documents draw a line between helping to produce a paper and occupying the author role on the paper. In the policies named here, AI may assist, but it does not belong on the author line.
Why accountability is the central reason
Nature makes accountability the key reason. Its editorial says authorship carries responsibility for the work, and it says AI tools cannot take that responsibility. That is a narrower point than saying AI never contributes anything. The claim is about the role authorship plays in scientific publishing.
ICMJE gives the practical version of that argument. Its recommendations say: “Because AI can generate authoritative-sounding output that can be incorrect, incomplete, or biased, humans are ultimately responsible for reviewing and ensuring the accuracy of any content generated with assistance of AI.” The ICMJE text adds that humans are responsible for citing appropriate sources, obtaining permissions when needed, and ensuring that plagiarism has not occurred when they use AI-generated content.
Springer Nature states the same idea in policy language. Its framework says accountability for scholarly content, evaluation, and editorial decisions cannot be delegated to AI systems. In other words, the system still looks for a human who can answer for the paper's claims, sourcing, and handling.
One reading of these documents is that authorship signals more than contribution. It signals an accountable role in a process that may include drafting, checking, revision, disclosure, correction, and response to challenges. Nature and ICMJE place that role with humans, not with a tool.
The steelman for the other side starts from the same texts. Nature says scientists had already used chatbots to help organize thinking, generate feedback, assist with writing code, and summarize research literature. Springer Nature says AI can assist clarity, efficiency, and exploration. If a tool materially shaped wording or analysis, some readers may think the most transparent approach would be to name it as an author. But the policies here reject that solution. They route transparency through disclosure instead.
That emphasis on a human answerable party also connects to other boundary questions this site covers, including whether AI can be an inventor, whether AI-generated work can be copyrighted, and whether AI can be morally responsible.
What journals still allow AI to do
The no-authorship rule is not a full ban on AI use. Nature says researchers who use large language model tools should document that use in the methods or acknowledgements sections, or in another appropriate section if the paper has no methods or acknowledgements. That means Nature's answer is not hide the tool; it is disclose the tool without treating it as an author.
ICMJE takes the same approach. Its recommendations say users should disclose which AI tool was used and for what purpose. The ICMJE text applies that expectation across manuscript preparation, editing, and review.
Springer Nature goes further and sorts AI uses by risk. Its low-risk examples include polishing language and suggesting structure. Its caution category includes drafting explanatory summaries, comparing results to existing literature, and recommending statistical tests or modelling approaches, but the policy says those uses require human oversight, verification, and disclosure. Its not-permitted category includes assigning authorship or accountability to AI systems, delegating peer review to a large language model, and generating hypotheses, analyses, or conclusions and presenting them as if they were human-derived.
Confidentiality is another limit. ICMJE says manuscripts submitted to journals are authors' privileged communication and warns that AI use in processing or evaluating them may violate confidentiality. Springer Nature states the same concern in operational terms: “Manuscripts, peer review reports, and sensitive data must not be shared with unsecured or public AI systems.”
So the rules in these documents are not anti-tool in a simple sense. They permit some assistance, sometimes extensive assistance, so long as the human role remains visible, responsible, and in control.
Why the question keeps coming up
Nature explains why this became a live question. Its editorial says chatbots can produce fluent language that is increasingly hard to distinguish from text written by people. Nature also says some scientists were already using chatbots as research assistants before ChatGPT reached a mass audience, and it notes that several preprints and published articles had already credited ChatGPT with formal authorship.
Nature also describes the misuse concern directly. Its editorial says: “The big worry in the research community is that students and scientists could deceitfully pass off LLM-written text as their own.” That concern fits the accountability logic. If readers cannot tell where text came from, and if the named authors try to blur responsibility, the author line stops doing part of its job.
There is also a broader trust problem. Nature says researchers should ask how transparency and trust-worthiness in knowledge production can be maintained when colleagues use software that works in a fundamentally opaque manner. Springer Nature's framework answers that concern by keeping judgement human and requiring disclosure. ICMJE answers it by keeping humans responsible for accuracy and sourcing.
That is why the answer remains no in the policies covered here. Nature, ICMJE, and Springer Nature do not deny that AI can shape drafts, summaries, wording, or exploratory work. They deny that such contribution is enough for authorship when the publication system expects a named author to carry accountability.
Frequently asked questions
Can I disclose AI use without naming the AI as an author?
Yes. Nature says researchers should document use of large language model tools in the methods or acknowledgements sections, or another appropriate section. ICMJE likewise says users should disclose which tool was used and for what purpose.
Do these policies ban all AI help with writing?
No. Springer Nature says AI may support clarity, efficiency, and exploration, and its examples include language polishing and structural help. ICMJE and Nature also allow AI use if humans review the output and disclose the tool's role.
Why not put the AI in the author list for transparency?
Nature and Springer Nature treat transparency as disclosure, not authorship. Nature says authorship carries accountability that AI tools cannot take, and Springer Nature says accountability for scholarly content and decisions cannot be delegated to AI systems.
Can peer reviewers use AI on submitted manuscripts?
Only with strong limits. ICMJE warns that using AI in processing or evaluating manuscripts may violate confidentiality, and it says editors, reviewers, or publishers should not upload submitted manuscripts into AI systems where confidentiality cannot be assured without authors' explicit permission. Springer Nature similarly says manuscripts and peer review reports must not be shared with unsecured or public AI systems.
Caveats: Nature's statements about scientists using chatbots as research assistants, about several preprints and published articles crediting ChatGPT with formal authorship, and about the risk of passing off AI-written text are single-sourced to its 24 January 2023 editorial. The broader no-authorship rule is also supported by ICMJE and Springer Nature policy texts.