A good peer review is a structured argument about whether a paper’s claims are supported by its evidence, delivered in a tone the authors can act on. It has a summary, major issues, minor issues, and a recommendation—and the quality of your reviewing directly shapes the literature you publish into.
Peer review is unpaid, slow, and often invisible work. It is also the mechanism that decides what counts as knowledge in your field. Reviewing well is how you build the kind of literature you want to read and cite.
What are the parts of a peer review?
Most journals expect the same basic structure. Following it makes the editor’s job easier and makes your review more useful to the authors.
-
Summary (1 short paragraph). Restate the paper’s main contribution in your own words. This does two things: it signals to the editor that you read the paper, and it shows the authors whether their contribution landed. If your summary does not match what the authors thought they claimed, that itself is valuable feedback.
-
General assessment (1 paragraph). State your overall judgment and the main reasons. Be specific: “the central claim is interesting but the experimental support is incomplete because…” beats “this paper has some issues.”
-
Major issues (numbered list). These are the problems that must be fixed for the paper to be acceptable—methodological flaws, missing baselines, unsupported claims, statistical errors, or related work that changes the interpretation. Each issue should be self-contained: state the problem, explain why it matters, and suggest a fix if you can.
-
Minor issues (numbered list). Typos, unclear figures, missing citations, terminology inconsistencies. These do not sink the paper but the authors will appreciate the catch.
-
Confidential comments to the editor (optional). Reserved for concerns about originality, ethics, or suitability that you do not want to share with the authors directly.
-
Recommendation. Accept, minor revision, major revision, reject. Most journals use a version of this scale.
How long should a peer review be?
Long enough to be useful, short enough to be readable. As a rule of thumb: 800–1500 words for a full research paper, with the major issues section carrying most of the weight. A review that is three sentences long is not a review; a review that runs to five thousand words usually has not prioritized.
The quality signal is not length but specificity. A single sentence that pinpoints a real methodological problem is worth more than a page of vague generalities.
How do you write constructive criticism in a peer review?
Constructive criticism is specific, actionable, and addressed to the work—not the people. A few rules:
- Attack the claim, not the author. “This experimental design cannot distinguish X from Y” is fair. “The authors do not understand experimental design” is not.
- Be specific. “Section 3.2 is unclear” tells the authors nothing. “The derivation between equations 4 and 5 skips a step; readers cannot verify the result without it” tells them exactly what to fix.
- Suggest fixes when you can. “This analysis is missing a control for Z” is more useful if you add “consider including the matched-pairs comparison.”
- Separate preference from defect. “I would have written this differently” is preference. “The claim is not supported by the data shown” is a defect. Only the second belongs in major issues.
- Assume good faith. Authors want their paper to be right. If something is wrong, the most likely explanation is a communication failure, not malice or incompetence.
How do you structure major versus minor issues?
The line is judgment, not definition. Ask: does this issue change whether the paper’s conclusions hold? If yes, it is major. If it is a clarity, formatting, or completeness issue that does not threaten the conclusions, it is minor.
When you are unsure, write it as major but flag your uncertainty. Editors appreciate a reviewer who admits they might be wrong more than one who pretends certainty they do not have.
How do you write the recommendation?
The recommendation follows from the major issues, not the other way around. Work out what would have to be true for the paper to be acceptable; then judge whether the authors can plausibly get there in a revision.
- Accept / minor revision: the central claims are supported by the evidence; remaining issues are fixable without new experiments or major rewrites.
- Major revision: the contribution is promising but the evidence or presentation is not yet sufficient. Be explicit about what a successful revision would look like.
- Reject: the claims are not supported by the evidence, the contribution is not novel, or the flaws are structural and cannot be fixed in revision.
What should you never do in a peer review?
A short list of failures that end reviewing relationships and occasionally careers:
- Demanding citation of your own work unless it is genuinely the most relevant source. Coercive citation is a documented ethical violation (see COPE guidance on citation manipulation).
- Stealing ideas. You see unpublished work; you do not get to use it. Confidentiality is the foundation of the entire system.
- Personal attacks. Reviewing is double-blind or single-blind for a reason. Tone matters, and ad hominem comments will be flagged by the editor.
- Reviews that are clearly AI-generated without reading. If you cannot engage with the specifics of the paper, decline the review.
Why good reviewing matters to you
Reviewing is the largest unpaid workload in academia, and it is also how you shape the field. More concretely: editors remember good reviewers. They invite them again, and they are the same editors who will eventually handle your submissions.
If you want a faster path to strong reviews of your own work, run a structured peer-style review on your manuscript before submission. Seeing your paper the way a reviewer would is the most efficient way to learn what reviewers actually check.
Frequently asked questions
How long should a peer review be? Most useful peer reviews run 800–1500 words for a full research paper, with the major issues section carrying most of the weight. The quality signal is not length but specificity—a single sentence pinpointing a real methodological problem is worth more than a page of vague generalities.
How do I write constructive peer review comments? Constructive criticism is specific, actionable, and addressed to the work rather than the authors. Attack the claim, not the person; be specific about where the problem is; suggest fixes when you can; separate your preferences from actual defects; and assume good faith—communication failures are more common than incompetence.
What are the parts of a peer review? A standard peer review has six parts: a summary of the contribution, a general assessment, major issues (numbered), minor issues (numbered), optional confidential comments to the editor, and a recommendation (accept, minor revision, major revision, or reject).
Can I disagree with the authors in a peer review? Yes, and you should. The point of peer review is to evaluate whether the claims are supported by the evidence. Disagreement is expected—what matters is that it is substantive, specific, and respectful, and that you make clear what would change your mind.
What should I never do as a peer reviewer? Never demand citation of your own work unless it is genuinely the most relevant source, never use ideas from the unpublished manuscript, never make personal attacks, and never submit a review you did not actually read the paper to produce. All of these are documented ethical violations under COPE guidance.