How to Disclose Your Use of AI in Research Manuscripts

Many researchers now use tools like ChatGPT and other artificial intelligence (AI) systems to support different stages of their manuscript preparation, from outlining to drafting to writing and revising. As these tools continue to evolve, you need to understand when and how their use should be disclosed in publications. In this article, you’ll learn why disclosure is needed, what current guidelines recommend, and practical ways to describe AI use clearly and transparently in your manuscripts.

Why you need to disclose your use of AI

AI and large language models (or LLMs) are great resources for writing, and their use is now ubiquitous across the sciences. Most people use these tools for some aspect of their work—brainstorming, rewriting, summarizing—even if they don’t mention it in their manuscripts. The challenge is that these tools can generate output that sounds confident and authoritative while being incorrect, incomplete, or biased. That means these tools carry real risks related to validity, confidentiality, transparency, and accountability. Because of those risks, you need to disclose how you use AI tools in your publications.

Some authors genuinely don’t know that they need to disclose AI use in their manuscripts. This may be because the disclosure guidelines have been evolving as AI tools (and their use) evolve, and not everyone is up to date on them. In January 2026, the International Committee of Medical Journal Editors (ICMJE) updated its guidelines on disclosing the use of AI in publications, and many journals follow these guidelines or adapt them to create their own policies.

Some authors are also afraid to admit they’re using AI for writing. They worry that disclosing AI use will make them look less skilled or that editors or reviewers will judge them poorly. But that’s not the case when AI is used responsibly. AI tools can be especially helpful if you’re writing in a language that isn’t your primary language, or if you need help rephrasing, structuring, or clarifying complex ideas. And when you use AI tools responsibly and disclose how you use them, it does not reflect poorly on you.

However, if you do not disclose your use of AI, you could face corrective actions or even be investigated for misconduct—and that reflects poorly on you. Transparency protects both you and your work, and disclosure is an essential part of responsible scientific and medical writing.

How to disclose your use of AI

According to the updated ICMJE recommendations, you need to disclose at the time of submission whether you used AI-assisted technologies in the production of submitted work. This includes LLMs, chatbots, or image generators.

You also need to describe what you used and how you used it, both in the cover letter with your submission and in the appropriate section of the manuscript. For example, if you used AI for data analysis or figure generation, disclose your use in the Methods section, like you would describe your use of statistical software to analyze data. Or if you used AI for writing assistance, disclose your use in the Acknowledgments section.

Authors often ask whether they need to disclose AI use if they only used it to check spelling and grammar. This use is generally considered acceptable and does not usually need to be disclosed. However, if you’re unsure or concerned, it’s always safe to disclose. If the journal editor thinks the disclosure is unnecessary, the related text can be removed before the article is published.

Another important thing to note is that AI-assisted tools should not be listed as authors. They do not meet all four ICMJE criteria for authorship, particularly around accountability for the accuracy, integrity, and originality of the work. Only humans can take responsibility for those aspects.

Most importantly, you—the human—are responsible for the submitted work, including any AI-generated content. AI tools are known to produce outputs that are incorrect, incomplete, or biased. So you must carefully review and edit everything they generate. You also need to ensure that you are not plagiarizing any material, including text or images produced by AI, and that you’re not inadvertently violating copyrights. That means you need to obtain permissions and be mindful of how you use AI-generated material.

Also, you cannot reference AI-generated content as a primary source. If AI gives you a claim, reference, or idea, you need to track down and verify the primary source and cite that source instead. And as always, you must appropriately attribute and fully cite all quoted material in the manuscript.

How much detail do you need to disclose

A common question is: how much information needs to be disclosed about AI use? Is it enough to simply say that you used an AI tool for writing assistance? Or do you need to give a detailed description?

The answer depends on what you used it for.

• If you used AI tools for the study itself (eg, data analysis), you need to describe your use of the tools in the Methods section. And you need to give enough detail that someone else could reproduce what you did. That means naming the tool, explaining how you used it, and providing any key settings or parameters that matter for interpretation.

• If you used AI tools for writing assistance, you need to disclose which tool was used and for what purpose. And this guidance applies to any stage of the writing process, from outlining to writing to editing to reviewing. For example, you might say that your team used a specific tool to develop an outline, improve clarity, or refine the wording in certain sections. And then you might add that your team verified all the content before submission. That last part—explaining that you checked and approved everything—can help reassure editors and reviewers that you exercised appropriate oversight.

If you want an extra layer of protection, you can track your use of AI tools so you have an audit trail. This tracking can be especially helpful if questions arise later about how the manuscript was prepared. You might track the tools, dates, prompts, inputs, and outputs. And because you need to fact-check everything, you might find it helpful to keep an annotated record of what you checked. The same approach can also be very useful when using AI tools for grant writing, where scrutiny can be just as high.

This might sound like a lot of work. But in practice, you can save prompts and outputs digitally, often in systems you’re already using to keep notes, drafts, or tracked changes. Also, once you build it into your workflow, the tracking doesn’t add that much extra effort—and it can be invaluable if anything ever comes into question.

Final thoughts

The ICMJE guidelines described in this article are widely accepted internationally, but some journals may have additional or slightly different policies. Before you submit, read the journal’s author instructions carefully and follow them closely. Doing so protects your manuscript and your credibility.