Learn · Peer review

How to peer review a paper, without becoming a poor reviewer

The risk in a review is not the flaw you miss. It is the problem you raise that is not one. Here is how to tell a real demand from a preference, and write a review you would sign with your name.

After this guide you will be able to:

The one distinction the whole review rests on

Before a point goes in your review, decide which of six things it is. Only the first two are demands the author must meet.

A point can beWhat it meansIs it a demand?
A methodological flawthe design or analysis cannot support the claimYes
A reporting deficiencythe work may be sound, but you cannot tell from the textYes
A legitimate disagreementa defensible choice you would have made differentlyNo, raise as a suggestion
A reviewer preferenceyour taste in methods, framing, or citationsNo, set it aside
A discipline mismatchthe wrong standard for this kind of paperNo, drop it
AI-generic criticisma checklist point that never touches this paperNo, delete it

Why the last four are the dangerous ones

A missed flaw is a normal limit of review, and the other reviewer or the editor often catches it. The damage comes from the other direction: dressing a preference as a requirement, applying the wrong field's standard, or pasting generic criticism that could apply to any paper. Each one costs an author real time on a change they did not owe, and it is the mark of a poor reviewer. So the discipline is not to find more. It is to demand only what is genuinely a flaw or a reporting gap, and to label the rest honestly as a suggestion.

Order it so the editor can act

An editor reads your review to make a decision. So order your points from fatal to editorial: the flaws that decide the paper first, the reporting gaps next, the suggestions last, each mapped to a specific line so the author can find it. A review that buries a fatal flaw under twenty style notes is hard to act on.

Keep the manuscript confidential

A paper under review is confidential. Do not paste it into a cloud AI to help you review, that shares an unpublished manuscript with a third party and can breach the confidentiality you agreed to. If you use any tool, use one that runs locally and sends nothing anywhere.

The self-check before you submit. Read your review back and ask of each point: is this a flaw, a reporting gap, or my preference? Would I make this demand if my name were on the review? Removing the points that fail that test is what separates a fair review from a harsh one.

Write the review you would sign.

You have a manuscript to review and a page of notes, and some of them are real flaws while others are just how you would have done it. Sorting them by hand, under a deadline, is how a preference slips through as a demand. The Reviewer's Notebook gives you the field guide and a local-first triage tool: classify each point into one of the six buckets, map it to a line, check it against the reviewer traps and the integrity questions, and export an ordered review and an editor summary. It structures your judgement. It does not review the paper for you, and it never leaves your computer.

See the Reviewer's Notebook · €39

Related guides

See also how to respond to reviewers, the author's side of the same exchange, and how to write a scientific paper. More guides are on the Learn page.