Using AI in peer review with confidence

AI tools are increasingly part of how research gets done, including in peer review

You are preparing a peer review report. You have read the manuscript carefully, shaped your assessment and drafted your feedback. Now you wonder whether an AI tool could help make your comments clearer or better organized. The practical questions are immediate: What tool can I use? What must stay confidential? If I use AI, do I need to say so?

Preparing to use AI in your review

A few quick questions can help you decide whether and how to use AI in a specific situation confidently:

• Is the AI tool private (it does not retain, reuse, share or learn from my input)?

• Am I certain that I am not sharing any text, data, figures or findings from the manuscript?

• Am I using AI to support my review, rather than replace my assessment?

• Do I need to disclose this use in my report?

Choosing the right AI tool

Elsevier’s updated policy states that reviewers should only use private AI tools to assist in the review process.

A private AI tool is one that does not retain, reuse, share or learn from the content submitted to it. Before choosing a tool, you are encouraged to check whether your institution provides an approved, enterprise, or locally hosted tool. These are more likely to include privacy protections, though this should still be verified rather than assumed. Free public tools are often not private by default.

For any tool being considered, you should look for clear statements that:

• user inputs are not retained beyond what is necessary to provide the service

• user inputs are not used to train or improve the AI model

• submitted content is not shared with third parties

• the tool permits content created or edited with it to be used for publication

One point applies regardless of which tool is used: the manuscript, or any part of it, should not be uploaded or pasted into an AI tool.

Using AI to support your review

There is genuine scope for AI to make the review process easier, provided confidentiality is maintained and you remain in control.

For example, you can use AI to improve the language or structure of your report, as long as you do not include confidential manuscript content. You can also use AI to support a general background literature search, provided you do not share unpublished details from the manuscript.

This might mean asking for help making your comments more concise, improving the tone of feedback, or organizing points more clearly. But the scientific assessment, interpretation and recommendation must remain your own.

"A reviewer's critical judgment and domain knowledge remain vital to evaluating science. At the same time, we're actively working on integrating AI capabilities into our own editorial systems to support reviewers in their work," explains Inez.

Disclosing AI use

Reviewers who use AI tools to help prepare their report, should include a short disclosure statement. Basic checks for spelling, grammar and punctuation do not need to be disclosed.

A disclosure could follow this format: “During the preparation of this report, I used [name of tool/service] in order to [reason]. After using this tool/service, I reviewed and edited the content as needed and I take full responsibility for my critical assessment and feedback on this manuscript.”

In summary

AI is becoming a familiar part of research, and peer review is no exception. Used thoughtfully with the right tools and clear disclosure, it can help reviewers focus their time and energy on what matters most: the quality of their assessment.

Read Elsevier’s updated generative AI policies for journals for the full guidance and FAQs.

Take this short quiz to assess your current understandingopens in new tab/window of the AI policy for reviewers!