How to Peer Review a Paper: Step-by-Step Guide + Template

To peer review a paper, read it twice: once for the contribution and any major flaws, once for detail. Then write a report with a short summary in your own words, numbered major comments tied to specific evidence, minor comments, and a recommendation that matches what you found. This guide covers each step, with a full template.

Most researchers learn how to peer review a paper by trial and error, because almost no one is taught. If an editor has just invited you (often because your name appeared on a related paper), that is a normal part of becoming a researcher, not a clerical mistake. What follows is the process experienced reviewers actually use, from invitation to submitted report.

Before you accept: a four-point checklist

Do not accept on reflex. Run through four questions first.

• Scope fit. Can you judge the core claims? You do not need to be expert in every method used, but you should be able to evaluate the main contribution. If you can only assess one part, tell the editor; they may still want you, paired with a second reviewer who covers the rest.
• Time. Journals typically ask for the report within a few weeks; the exact deadline is in the invitation. A careful first review often takes several hours spread over multiple sittings. If your next month is already full, decline now rather than late.
• Conflicts of interest. Recent co-authors, current or former supervisors, close collaborators, direct competitors on the same question, or anything that could look like a conflict: disclose it and let the editor decide.
• Confidentiality. By accepting, you agree to keep the manuscript confidential. You cannot share it with colleagues or students without the editor's permission, cite it, use its ideas, or upload it anywhere. More on that below.

Declining is fine and editors expect it. A fast "no, but try these two colleagues" helps them more than a slow yes that turns into a late report.

How to peer review a paper: the two-pass reading method

Reviewing a journal article in one linear read is the most common beginner mistake: you end up with forty notes about commas and no view of whether the paper should be published. Read it twice, with a different job each time.

Pass 1: contribution and major flaws (skim, then think). Read the abstract, introduction, figures, and conclusions. Ignore typos. Answer three questions in writing:

1. What do the authors claim is new?
2. Does the evidence in the paper support that claim?
3. Is there a flaw that no revision could fix: wrong method for the question, missing controls, conclusions the data cannot carry?

If pass 1 reveals a fatal problem, you do not need three more hours of line edits. Your report can be short, clear, and centered on that problem.

Pass 2: the detailed read. Now go section by section. Check that methods match what the results claim, that figures show what the text says they show, that statistics are appropriate, that limitations are acknowledged, and that the writing is clear enough to follow. Take notes with section, page, or line numbers as you go; these become your comments.

The two-pass method also explains something authors often miss: reviewers form their judgment of the contribution early, from the framing and the figures, before they ever reach the details.

How to write a peer review report

Nearly every good review, in any field, has the same four parts, in this order.

1. Summary. One paragraph restating the paper's question, approach, and main claims in your own words, not copied from the abstract. This shows the authors and editor that you understood the paper; if your summary and the authors' intent diverge, that gap is itself useful feedback.
2. Major comments. Numbered points that affect the validity or interpretation of the work: unsupported claims, methodological gaps, missing analyses of existing data, overstated conclusions. Each one should name the exact location, explain the problem, and say what would resolve it.
3. Minor comments. Numbered, briefer points: unclear sentences, figure legibility, notation, missing definitions, typos worth flagging.
4. Confidential comments to the editor. A separate box the authors never see. Use it for your honest bottom line, any suspicion of ethical problems (plagiarism, image manipulation, duplicate submission), and any limits of your own expertise.

Numbering matters more than it looks. Authors respond point by point in their revision, and editors check that response against your list; see how that works from the author's side in our guide to responding to reviewer comments.

Peer review template you can copy

Use this peer review template as your starting structure and adapt it to the journal's form.

MANUSCRIPT: [Title / ID]
RECOMMENDATION: [Accept / Minor revision / Major revision / Reject]

SUMMARY
In 3–5 sentences, restate the research question, the approach, and
the main claims in your own words. Note briefly what the paper
contributes if the claims hold.

MAJOR COMMENTS
1. [Location: section/page/line.] [What the problem is, with the
   specific claim and the specific evidence at issue.] [What would
   resolve it: a reanalysis, a qualified claim, a clarification.]
2. ...
3. ...

MINOR COMMENTS
1. [Location.] [Brief issue: unclear sentence, figure labeling,
   missing definition, inconsistent terminology.]
2. ...

CONFIDENTIAL COMMENTS TO THE EDITOR
[Your candid overall assessment. Any ethical concerns. Any parts of
the paper outside your expertise. Nothing here that contradicts the
comments the authors will see.]

One rule ties the template together: nothing in the confidential comments should surprise an author who has read your public comments. Editors distrust reviews that say "fine work" to the authors and "reject" to the editor.

Accept, minor, major, or reject: how recommendations map to comments

Your recommendation is not a separate judgment; it should follow mechanically from the comments you wrote.

If you recommend "major revision" but your comments are all minor, or "reject" without an unfixable major comment, revise one or the other until they agree. Editors read the comments first and the recommendation second; a mismatch weakens both.

Working through this list has a side effect worth noticing: it is exactly the checklist your own manuscripts face, and a pre-submission review with Thesify Reviewer applies the same reviewer's-eye reading to your own draft before an editor ever sees it.

Tone: critique the work, not the authors

The difference between a review that improves a paper and one that starts a feud is almost entirely tone and specificity. (The quoted phrasings in this section are composite examples based on common feedback patterns, not quotes from real reviews.)

• Address the work. "The manuscript does not report how participants were recruited," not "the authors carelessly omit their recruitment procedure."
• Be specific and evidence-based. Vague verdicts are unusable. Weak: "The results are unconvincing." Strong: "Figure 2 shows overlapping confidence intervals between conditions, which does not support the claim of a significant difference in lines 142–144. Please either report the direct statistical comparison or qualify the claim."
• Say what would fix it. Every major comment should point to a resolution, even if the resolution is "qualify the conclusion."
• Do not demand citations to your own work. Suggest a reference only when the paper genuinely cannot stand without it, and never as a way to raise your own citation count. Editors notice.
• Note strengths honestly. One or two sentences on what the paper does well tells the editor which parts to protect in revision.

For a library of weak-versus-strong phrasings, see our collection of reviewer comment examples.

What not to do as a peer reviewer

Three mistakes cause most first-review disasters.

Do not upload the manuscript to public AI tools. The manuscript is confidential and unpublished. Pasting it into a public chatbot or uploading the PDF to a web tool shares it with a third party, and publishers broadly prohibit this in their reviewer policies. Using AI on someone else's confidential manuscript is a different act from an author running an AI peer review tool on their own draft before submission, which breaches no confidence. For where the line currently sits, see our overview of AI in peer review.

Do not fish for new experiments beyond the paper's claims. Ask whether the evidence supports the claims made, not whether the paper answers every question you find interesting. "This would be stronger with a two-year longitudinal follow-up" is a request for a different paper. If a claim genuinely needs more evidence, the authors can also fix it by narrowing the claim.

Do not review past your expertise silently. If the paper's statistics, a specialized method, or a subfield section is beyond you, say so in the confidential comments. Editors routinely assemble reviewers with complementary expertise; pretending to cover everything gives the editor a false picture of what has actually been checked.

FAQ

How long does a peer review take?

Plan for several hours of focused work for a first review, spread over multiple sittings, depending on the paper's length and how close it sits to your expertise. Journals usually set a deadline of a few weeks. Experienced reviewers get faster, mostly because the two-pass method becomes automatic.

Can PhD students peer review papers?

Yes. Editors invite PhD students who have published in the relevant area, and supervisors sometimes involve students in reviews, but only with the editor's explicit permission, since the manuscript is confidential. Co-reviewing with your supervisor, disclosed to the journal, is a common and legitimate way to learn.

Do peer reviewers get paid?

At the large majority of journals, no: reviewing is unpaid service to the field. Some journals offer small honoraria, discounts, or formal recognition, but these are the exception. The practical returns are early exposure to new work, a sharper eye for your own writing, and a working relationship with editors.

Can I use AI to help write my review?

Not on the manuscript itself: uploading a confidential submission to public AI tools is broadly prohibited by publisher policies, and some journals restrict AI use in reviewing altogether; check the journal's reviewer guidelines before using any tool. Using AI to polish the grammar of your own already-written comments is treated differently by different journals, so the same check applies.

Before you submit, see your paper the way reviewers will. Thesify Reviewer runs a pre-submission review of your manuscript. It flags the issues peer reviewers are most likely to raise, from unsupported claims to methodological gaps, before a journal ever sees it. Try Thesify Reviewer