Skip to content

Reviewer norms and the agreement banner

This page exists so that anyone — author, reviewer, or editor — has a stable URL to consult for the exact text of the agreement that AI4Meta shows to reviewers when they open a shared project link.

It also restates the relevant peer-review norms from COPE and ICMJE, in plain language, with citations. If you are a reviewer who arrived here by clicking the agreement banner, this is the document that explains exactly what you just agreed to.

The agreement banner

When a reviewer opens a share link for the first time, AI4Meta shows the following banner. The reviewer must click I agree and continue before the project content loads.

You are about to view material shared with you in confidence.

The content of this AI4Meta project — including codebook, search strategy, screening decisions, extractions, and analyses — is the work product of the project authors. It has been shared with you for the purpose stated in the share invitation, typically peer review or methodological audit.

By continuing, you agree to the standards established by the Committee on Publication Ethics (COPE) and the International Committee of Medical Journal Editors (ICMJE) for peer review:

  1. You will treat the material as confidential. You will not show, discuss, or quote it to anyone other than the editor who invited you, except where authorized in writing.
  2. You will not use this material to advance your own work. You will not adapt the codebook, replicate the search strategy, or build on the extracted dataset for a project of your own without explicit written permission from the project authors.
  3. You will declare any conflicts that would prevent impartial review.
  4. After your review is complete, you will delete or destroy any local copies of the material.

Unauthorized reuse of material seen in peer review is, by every major journal's policy, research misconduct. AI4Meta records your access to this share link, and the project authors retain a cryptographic timestamp of the project state on the Bitcoin blockchain via OpenTimestamps.

I agree and continue / Decline

Clicking Decline closes the link without loading content. Clicking I agree and continue records your agreement (with timestamp) and loads the project at the share link's scope.

What COPE and ICMJE actually say

The text of the banner is grounded in two widely-cited bodies of guidance.

COPE Ethical Guidelines for Peer Reviewers

Relevant excerpts (full text at publicationethics.org/peerreview):

  • "Reviewers should respect the confidentiality of peer review and not reveal any details of a manuscript or its review, during or after the peer-review process, beyond those that are released by the journal."
  • "Reviewers should not use information obtained during the peer-review process for their own or any other person's or organization's advantage, or to disadvantage or discredit others."

This is the primary norm. A reviewer who copies a codebook from a manuscript-under-review, even years after the review, has violated COPE guidelines and is subject to sanction by any journal that subscribes to them — which is most.

ICMJE Recommendations

The ICMJE Recommendations (full text at icmje.org) extend the COPE guidance with specific language for biomedical journals:

  • "Reviewers must keep manuscripts and the information they contain strictly confidential."
  • "Reviewers must not publicly discuss authors' work and must not appropriate authors' ideas before the manuscript is published. Reviewers must not retain the manuscript for their personal use and should destroy paper copies of manuscripts and delete electronic copies after submitting their reviews."

ICMJE is the standard for biomedical journals (NEJM, JAMA, BMJ, Lancet, all member journals). A reviewer who breaches these norms while reviewing for an ICMJE-member journal is subject to that journal's misconduct procedure, which routinely results in retraction of any subsequent paper found to have copied the under-review material, plus a permanent ban from reviewing for the journal.

What this means in practice

For authors: when you share a project with a reviewer through AI4Meta, the reviewer is on notice. The agreement is logged. The norms are not aspirational; they are enforceable through the journal in question.

For reviewers: do not use what you see in a shared AI4Meta project to inform your own competing work. If reading the project gives you ideas for your own research that you cannot un-think, the standard remedy is to decline future review invitations on related topics for a reasonable period (typically 1 year) and to avoid initiating any directly competing project until the manuscript under review is published or formally withdrawn. If the overlap is unavoidable, disclose it to the editor.

For editors: the agreement banner and the access log are designed to be admissible evidence in a misconduct investigation. If a complaint is filed and you need the access record, contact AI4Meta support; we will provide the timestamped access log on receipt of a journal-issued request.

Why we surface this to reviewers

Most reviewers behave well. The norms above are widely known and widely respected. The reason AI4Meta interrupts the reviewer flow with an explicit click-through is not because we expect reviewers to misbehave, but because the act of clicking through the agreement makes "I didn't know" indefensible in any subsequent dispute. It is a small piece of friction with a large evidentiary payoff.

The same reasoning informs AI4Meta's other protections: pre-registration is friction (you have to write the protocol up front), scoped share links are friction (you have to pick a scope), OpenTimestamps stamping is friction (it adds a few seconds to lifecycle events). Each piece of friction trades a small amount of author and reviewer time for a large amount of evidentiary clarity if a dispute arises.

See also