Why can an author give a paper to AI when a reviewer cannot?

Review services

Many AI services are available that enable you to get a review of your paper. I am not talking about general Large Language Models. I am referring to specific products that are aimed at research scholars. I see adverts for these on my social media feeds almost every day.

The same paper, a different rule

Imagine that I have finished writing a research paper. Before submitting it, I upload it to an AI service and ask it to identify weaknesses, unsupported claims and questions a reviewer might raise. I examine the feedback, make the changes that I think are justified and submit the paper to a journal.

My first question is whether this use complies with the policy of the journal I intend to submit to. If the paper is rejected, I would also need to check whether that earlier AI use meets the next journal’s requirements.

Now imagine that the journal sends my research to you for peer review. You upload it to the same AI service and ask it to identify the same kinds of problems, intending to assess its suggestions before writing your report.

If you are reviewing for an Elsevier journal, uploading that manuscript is prohibited. The paper and technology may be identical, but your responsibilities are different.

Whose manuscript is it?

Elsevier’s current author policy permits AI assistance during manuscript preparation, subject to human oversight, disclosure and safeguards for privacy and intellectual property. Authors must check the service’s terms, including what rights it obtains over uploaded material and whether it can use that material for training. Permission to use AI therefore comes with conditions.

Reviewers occupy a different position. They have been entrusted with confidential work, and Elsevier prohibits uploading submitted manuscripts, or information from them, into AI tools. Its explanation expressly includes services described as private. The statement on their web site says:

“Reviewers should not upload a submitted manuscript or any part of it into an AI tool as this may violate the authors’ confidentiality and proprietary rights and, where the paper contains personally identifiable information, may breach data privacy rights.”

This is understandable. An author may be authorised to decide how a manuscript is processed, subject to obligations to co-authors, institutions, funders and research participants. A reviewer cannot assume the same authority simply because a journal has supplied a copy.

Receiving a manuscript for review gives you permission to assess it within the journal’s rules. It does not give you unrestricted permission to send it elsewhere.

When the tool gives different advice

The situation becomes more complicated when an AI provider offers guidance that researchers could reasonably interpret as instructions for conducting a journal review. Jenni provides a useful example because its website addresses both authors preparing papers and reviewers assessing them.

One Jenni guide promotes simulated reviewer feedback on an author’s own manuscript before submission. That is consistent with the distinction above: authors use feedback to strengthen their work while retaining responsibility for the decisions they make.

However, its separate guide to writing a peer-review report includes the instruction: “Step 1: Upload the manuscript PDF to your Jenni library”. The subsequent steps recommend analysing claims and citations, running its Reviews feature and using the output as a starting point for the report.

For someone reviewing a confidential manuscript for Elsevier, that workflow conflicts with the publisher’s policy. Writing the final report yourself does not resolve the problem, because the prohibited action occurs when you upload the manuscript.

Jenni’s guide acknowledges that disclosure requirements vary and warns against presenting an AI-generated review as your own expert opinion. But these qualifications address different issues from whether you have permission to disclose the manuscript to an external service.

Does private mean permitted?

Researchers may understandably assume that a service offering privacy protections is suitable for confidential work. But the service’s data protections and your authority to upload a document are separate questions.

Jenni’s documentation says that drafts, prompts and outputs are not used to train its own models. It also explains that an AI feature may send relevant document content to a third-party service when needed to complete a request. That is, it says (but see their privacy page for complete information):

“Jenni sends only the content needed to complete the AI action you requested.”
“Third-party services may process relevant content when needed to provide an AI feature.”

Elsevier’s restriction makes this distinction explicit. Even using a private tool does not provide reviewers with permission to upload submitted manuscripts, although limited supporting uses of AI are allowed where confidentiality is maintained.

What happens after rejection?

Suppose an author uses an AI service to obtain feedback, submits the revised manuscript to one journal and receives a rejection. Does that earlier AI use prevent submission to an Elsevier journal?

The policy does not establish an automatic prohibition of that kind. The relevant questions concern whether the use complied with its conditions, whether rights and confidentiality were protected, and whether the required disclosure is made. We should therefore avoid suggesting that using Jenni necessarily closes the door to an Elsevier journal.

I would welcome clarification from publishers on how they assess AI assistance used before a manuscript reaches their journal.

Nevertheless, the example reveals a practical burden. Researchers may need to consider potential submission requirements when choosing a tool, rather than discovering after using it that its terms create difficulties for publication.

How would anybody know?

There is also the question of enforcement. Suppose a reviewer uploads a manuscript, reads the AI assessment and then writes a report without disclosing what they have done.

The report alone may not establish that the manuscript was uploaded. Unless other evidence emerges, the editor may have little opportunity to identify the breach.

This does not make a confidentiality rule pointless. Peer review already depends on trust: reviewers are expected to protect unpublished findings and respect restrictions on sharing material, even when nobody can routinely monitor their conduct.

But a policy needs more than a place on a website. Reviewers need to understand the restriction when accepting an invitation, and AI providers should explain that a tool’s capabilities do not establish permission to use it on confidential manuscripts.

Rules researchers can follow

Researchers already navigate substantial variation in journal requirements. AI adds questions about their role, the information they can upload, the task they want help with and the disclosures they must make.

Even the apparently simple question, “Can I use AI for peer review?”, is insufficient. Improving the wording of a report, searching background literature and uploading a manuscript for analysis involve different activities and different confidentiality risks.

Publishers and AI providers could make those distinctions much clearer. Guidance should explain acceptable uses through concrete examples and place confidentiality warnings alongside instructions that involve uploading documents.

Greater consistency across publishers would also help, without requiring every activity to be treated identically. Shared principles could distinguish author preparation, reviewer assistance and scientific assessment, while explaining which safeguards apply in each situation.

The difference between an author and a reviewer is defensible. What needs attention is how clearly that difference is communicated, particularly when a service shows researchers how to perform an activity that a journal prohibits.

Researchers remain responsible for following the rules attached to a manuscript. But publishers and tool providers also have a responsibility to make those rules understandable before the upload happens.

This leaves me with two further questions. Publishers and journals issue the rules, but should they take greater responsibility for helping researchers understand and follow them? Clear(er) explanations when a reviewer accepts an invitation would be a useful starting point.

And how should compliance be assessed? If breaches are difficult to detect, publishers need to explain how concerns can be reported and investigated, while recognising that professional trust will remain essential.

I publish two free newsletters each week

Governance & Research in HE
A weekly brief for higher education leaders and researchers examining governance, leadership and research in practice.
https://buff.ly/GhsZqVl

Publishing with Integrity
Examining scholarly publishing through an ethical lens, challenging assumptions and supporting academic career development.
https://buff.ly/WiG1yFb

If you subscribe, you will be alerted whenever I publish a new edition.

About the author

Professor Graham Kendall is Vice-Chancellor of GlobalNxt University, Malaysia, and Emeritus Professor at the University of Nottingham. He has published more than 300 peer-reviewed papers and served as an editor-in-chief and associate editor of academic journals. He writes about university governance, leadership, research and integrity in scholarly publishing.

Archived URls

The links in this article have been archived, just so that we can be sure we are looking at the same information.

Publishing with Integrity

Review services

The same paper, a different rule

Whose manuscript is it?

When the tool gives different advice

Does private mean permitted?

What happens after rejection?

How would anybody know?

Rules researchers can follow

I publish two free newsletters each week

About the author

Archived URls

Elsevier’s Generative AI policies for journals

Jenni.ai

Jenni Author Guidelines for Peer Review

Jenni Reviewer Guidelines

Jenni Privacy Page

Originally published on LinkedIn

Would you like to know when the next edition is published?