Can You Upload Reviewer Comments to an AI Tool?

Can Reviewer Comments Be Pasted Into AI?

You receive a decision letter containing two pages of reviewer comments. Some are straightforward. Others are cryptic, contradictory, or ask for substantial revisions. An AI tool could organize the comments, translate them into a revision checklist, suggest how to respond, or help polish your response letter.

Can you paste the comments into the tool?

The answer depends partly on whose comments they are, why you possess them, what information they reveal, and what the journal permits. An author using comments received about their own manuscript is not in exactly the same position as a reviewer uploading the confidential review they are preparing.

Sometimes, but First Determine What the Comments Reveal

"Reviewer Comments" Can Mean Several Different Things

The author's copy of reviewer comments is different from the reviewer's working material

Suppose an author receives comments such as, "The rationale for choosing this sampling strategy needs clarification." The author may want AI assistance interpreting the request or organizing a response.

Now consider the reviewer who originally wrote that comment. The reviewer has access to the submitted manuscript, unpublished findings, editorial correspondence, perhaps confidential comments to the editor, and information obtained solely through the review assignment.

Those two situations should not be collapsed into a single rule.

Reviewer comments can reveal manuscript content

A review report does not need to reproduce the manuscript word for word to disclose it. Reviewer comments may summarize unpublished findings, quote passages, identify methodological weaknesses, discuss novel methods, describe datasets, or reveal arguments that have not yet been published.

For example, a comment saying, "The authors' finding that intervention X reduced outcome Y by 42% needs sensitivity analysis" reveals more than the reviewer's opinion. It also reveals an unpublished result.

That is why apparently harmless review text may still qualify as confidential peer-review material.

Journal policies can distinguish supportive AI use from disclosure of manuscript content

Current publisher policies are not identical. Researchers should therefore check the journal that actually handled the manuscript.

ICMJE states that reviewers must follow the journal's AI policy or obtain permission before using AI to facilitate a review. It also emphasizes that manuscripts are confidential and that uploading them to AI systems may be prohibited where confidentiality cannot be assured.

Elsevier's current journal policy permits reviewers to use AI in a supportive capacity, such as improving the language or structure of a review report or assisting with background literature searches, provided confidentiality is maintained and human oversight remains. However, reviewers must not upload the submitted manuscript or information from it into an AI tool. Elsevier also requires disclosure of qualifying AI assistance in the review report, while basic spelling and grammar checks do not require disclosure.

These distinctions matter. "AI is allowed for reviewers" does not mean "the manuscript may be pasted into AI."

Authors responding to reviews face a different decision

An author receiving reviews of their own manuscript is generally not exercising the reviewer's confidential access to somebody else's unpublished work. Even so, the decision letter can contain material that deserves protection.

It may include reviewer identities in an open or signed review system, confidential editorial correspondence, manuscript identifiers, submission-system links, editor names and contact details, or text that reproduces unpublished manuscript content. The journal may also impose its own conditions on AI-assisted manuscript preparation.

If the task is simply to convert ten reviewer requests into a checklist, you may not need to upload the entire decision letter.

Reviewer identity deserves particular care

Peer-review models vary. In single-anonymized or double-anonymized review, reviewer identities may intentionally be withheld from authors. In signed or open review, identities may be disclosed under the journal's model.

Do not attempt to use AI to infer or uncover an anonymous reviewer's identity from writing style, citations, methodological preferences, or other clues. That would undermine the journal's review model and could conflict with confidentiality expectations.

If a reviewer is named, their name is still personal information that may be unnecessary for the AI task. Removing it before asking for help interpreting a methodological comment is a straightforward form of data minimization.

Editorial correspondence can contain more than reviewer comments

A decision email may include editor-only material that has been forwarded to the author, internal identifiers, submission links, editorial judgments, reviewer recommendations, or other contextual information. Uploading the complete email can therefore disclose much more than the comments you wanted help understanding.

This is another reason to distinguish the information you need from the container in which you received it. Copying a relevant non-sensitive comment may be materially different from uploading the entire correspondence file.

Using AI to draft a response does not transfer responsibility

AI can help reorganize a response letter, identify where several reviewer comments overlap, or suggest a clearer explanation. It cannot determine whether the reviewer is scientifically correct merely because its answer sounds decisive.

A model may confidently recommend an unnecessary analysis, misunderstand disciplinary conventions, invent a methodological requirement, or produce a conciliatory response that agrees to something you should instead challenge politely.

The authors remain responsible for deciding which revisions improve the paper, which requests require clarification, and which comments warrant a reasoned disagreement.

Do not confuse comments received as an author with comments written as a reviewer

This distinction becomes particularly important when researchers play both roles regularly. A review report you receive about your own paper and a review report you are preparing about someone else's paper arise from different relationships.

If you are acting as a reviewer, your access exists because the journal entrusted confidential material to you for a defined purpose. The restrictions surrounding confidential peer-review material and AI should therefore govern the decision.

The AI service's data practices still matter

Even where the journal permits a particular AI-assisted task, the service you choose matters. Review text entered into an AI system may be processed, retained, logged, or used according to the provider's applicable terms and settings.

A journal permitting supportive AI use does not automatically approve every public chatbot. Researchers should still consider what happens to research information submitted to the AI service.

An Author Wants AI Help Organizing Reviewer Requests

The useful AI task required substantially less information than uploading the entire peer-review correspondence.

Common Mistakes With Reviewer Comments and AI

First Identify Which Side of Peer Review You Are On

The easiest way to avoid overgeneralizing is to begin with your role. Authors, reviewers, and editors possess peer-review information under different circumstances.

For reviewers in particular, the safer starting point is conservative because the unpublished research belongs to somebody else and was disclosed for a limited editorial purpose.

Before Entering Reviewer Comments Into AI

Questions About Reviewer Comments and AI

Reviewer Comments Are Not Automatically Ordinary Text

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.