Scope: Structure · tone · ethics · timingData: Editorial standards + COPE guidelines + journal reviewer instructionsLast reviewed: March 2026Source: Manusights editorial team (researchers with publications in Cell, Nature, Science)Cite this guide ↓

How to Write a Peer Review: A Practical Guide for Researchers

Reviewing papers is one of the most underrated career moves in academic science. It's how you learn what editors look for, build relationships with journals before you submit your own work, and understand why good papers get rejected. Most PIs start reviewing in their postdoc years. Some PhD students start in year 3 or 4 with their supervisor's guidance. This guide covers the mechanics of writing a good review, the ethics involved, and what not to do.

Sources are cited in the references section below.

Why Review, and When to Start

Peer review isn't charity work. It's a strategic career activity with tangible returns.

Learn what makes papers succeed

Reading 10-20 manuscripts a year, including the ones that don't work, teaches you more about scientific writing than any course. You'll start seeing patterns in what gets rejected and what gets through.

Build editorial relationships

Editors track reviewer quality. A reliable reviewer who delivers thoughtful, on-time reviews gets remembered. When you submit your own papers, that reputation matters.

Join journal reviewer lists

Your name enters the journal's reviewer database. This is searchable by editors across the publisher's portfolio. One good review for a Nature journal can lead to invitations from other Nature titles.

Be seen as a contributor

Reviewing shows you're engaged in the field beyond your own publications. Search committees and grant reviewers notice peer review activity on CVs under Professional Service.

When to start and how much to take on

  • • Most journals start inviting postdocs to review once they have 2–3 first-author publications in the relevant area
  • • PhD students can review with explicit permission from their PI. This is ethical and common as long as the PI supervises and takes responsibility
  • • A thorough review takes 4–8 hours for a typical original research article, less for shorter papers or follow-on revisions
  • • 2–5 reviews per year is a reasonable load for a postdoc; more than 8–10/year starts affecting your own work

Getting Invited to Review

Most junior researchers wait for invitations to appear in their inbox. That works eventually, but there are faster paths.

Build a publication record first

You need at least 1-2 papers in the area before journals will invite you. Editors search for reviewers by matching manuscript keywords to author publication records. No publications in the topic area means you won't appear in those searches.

Register on Publons (Web of Science reviewer recognition)

Create a profile and note your expertise areas. Some editors search Publons directly when they can't find reviewers through the usual channels. It also gives you a verified record of your review activity for your CV.

Submit your own papers

When you submit to a journal, editors see your email domain, research area, and recent publications. You'll start getting invitations naturally from journals you've submitted to, even if your paper was rejected.

Email a section editor directly

If you want to review for a specific journal, write to the managing editor or a section editor. Introduce yourself, list your expertise and recent publications. Keep it to 3 sentences. Many journals accept reviewer self-nominations.

Ask your mentor to suggest you

Your PhD supervisor or postdoc mentor can directly suggest you as a reviewer when they decline an invitation. This is how most junior researchers get their first review invitations, and it carries the mentor's implicit endorsement.

The Structure of a Peer Review

Most journals use a structured submission form with text boxes, but always write the narrative review in a document first, then paste in. This gives you a record and lets you edit before submitting.

1. Summary paragraph (3-5 sentences)

Restate the paper's central claim and what the authors did to support it. This proves you read it carefully and helps the editor understand whether your concerns are fundamental or minor. Don't evaluate the paper here. Just show you understood it.

2. Major concerns (numbered)

Issues that must be addressed for the paper to be publishable. These are problems with the science itself: missing controls, unwarranted conclusions, statistical errors, missing comparison to prior work, insufficient sample size, or methodology that doesn't support the claims. Each concern should be specific and actionable.

3. Minor concerns (numbered)

Presentation issues, unclear figures, citation gaps, or language corrections that don't affect the scientific conclusions. These are things the authors can fix without new experiments or major rewriting.

4. Recommendation

Major revision, minor revision, or reject. Be honest and match your recommendation to your concerns. If you listed five major concerns that each require new experiments, don't recommend minor revision. If your only concerns are about figure clarity and one missing citation, don't recommend major revision.

What NOT to include

  • • A list of typos. That's copy-editing, not peer review
  • • Complaints about writing quality without specific examples
  • • Vague comments like "the conclusions are not supported" without explaining why

Length: a useful review is typically 400–800 words. Longer is not better. Every comment should be actionable.

Major vs. Minor: Getting the Calibration Right

This is where junior reviewers mess up most often. They call everything “major,” or they bury a fatal issue inside the minor comments. Editors are reading your review partly to judge whether you know the difference.

Usually major

  • • The main claim depends on a control that is missing or inadequate
  • • The statistics do not support the central conclusion
  • • The paper overclaims beyond what the data can actually show
  • • A key comparison to prior literature is missing and changes the novelty story
  • • The methods are too unclear to assess reproducibility
  • • A requested experiment would genuinely decide whether the conclusion is true

Usually minor

  • • Figure labels are unclear or abbreviations are undefined
  • • One or two citations are missing
  • • The discussion needs tightening but the science is sound
  • • A schematic model would improve readability
  • • The methods need small clarifications, not a redesign
  • • Language is awkward in places but understandable
Useful rule: if fixing the issue would change whether the paper is publishable, it is major. If fixing it would mostly improve clarity, presentation, or completeness, it is minor. Not every extra experiment is a major concern. Some are just things you would personally like to see.

Confidential Comments to the Editor

Many systems give you a separate box for comments to the editor. Most of the time, you do not need to say anything there that you would not say to the authors. But there are a few situations where that box matters.

1.You think the paper is fundamentally unsound but do not want the authors to see a blunt recommendation line like 'this should be rejected'
2.You suspect image manipulation, plagiarism, salami slicing, or another integrity problem
3.You have a mild conflict or concern about overlap with your own work that the editor should know
4.You want to explain your recommendation briefly in editor-facing language: for example, 'the work is interesting but the core causal claim is not supported without the missing control in Figure 3'

What that box is not for

It is not a place to say one thing to the editor and a softer version to the authors. Editors notice that immediately. If your public comments say “promising paper, minor concerns” and your confidential note says “reject,” you look untrustworthy. Use the confidential box for context, not for a second hidden review.

Writing Comments That Are Actually Useful

The difference between a review that helps authors improve their paper and one that frustrates everyone comes down to specificity.

Good review comments

  • +"The authors should validate the finding in an independent dataset" — specific and actionable
  • +Include citations when you know relevant prior work the authors missed. Say what was published, don't just say "this has been done"
  • +Frame concerns as questions when appropriate: "It's unclear whether the effect persists in the absence of X — have the authors tested this?"
  • +If you think a paper should be rejected, say so clearly with the specific reason
  • +When you disagree with a method, acknowledge whether it's a preference vs. a scientific problem

Unhelpful review comments

  • -"Validation is needed" — vague, doesn't say what kind or why
  • -"This has been published before" without citing what was published
  • -"The methods are inadequate" without specifying which methods or what's wrong
  • -Writing a detailed 3-page review of a paper you think is fundamentally flawed instead of recommending rejection with a clear reason
  • -Sarcastic or condescending tone that editors will flag as unprofessional

Tone matters. Your comments will be read by the authors (anonymously) and by the editor. Both audiences matter. Professional, clear, and constructive beats clever or cutting.

Common Reviewer Mistakes

xRequesting experiments that would take 2 years. Ask whether the conclusion is supported by the current data, not whether every possible experiment has been done
xRecommending rejection without explaining what would make the paper publishable in any venue
xRevealing your identity in the review. Don't thank the authors for "your great work" if you're double-blind
xUsing the review as a venue to promote your own papers by requesting citations to your work that aren't genuinely relevant
xReviewing a paper where you have a clear conflict: competing paper under review, close collaborator as author, personal animosity toward an author
xSitting on the review for 6 weeks. Late reviews delay publication and frustrate editors, authors, and other reviewers who delivered on time

Ethics and Conflicts of Interest

Peer review runs on trust. The manuscript you receive is not public, and the review process imposes specific ethical obligations.

Confidentiality

The manuscript is not public. Don't share it, don't discuss it with colleagues (unless your journal explicitly allows collaborative reviews), and don't use ideas or data from it before the paper is published. Violating confidentiality is one of the most serious ethical breaches in peer review.

Conflicts requiring recusal

  • • You have a paper competing with this one under review or in preparation
  • • You've recently collaborated with one of the authors (typically within the last 3 years)
  • • You have a financial interest in the outcome
  • • You have a significant personal conflict with an author

When in doubt, disclose

Email the handling editor and describe the potential conflict. They'll decide whether you can still review. If you must decline, do it promptly (within 3–5 days) and suggest an alternative reviewer if you can.

Using AI to assist with peer reviews

Check the journal's policy before using any AI tools. Many journals (Nature, Cell, NEJM) prohibit using AI tools to write reviews because the manuscript is confidential and cannot be shared with third-party systems. This is an evolving area. Always check the specific journal's current policy before uploading any manuscript content to an AI system.

Reviewer Recognition

Peer review has historically been invisible work. That's changing. Several platforms now let you document and get credit for your reviews.

Publons (Web of Science Reviewer Recognition)

The main platform for recording and verifying peer review contributions. Most major publishers (Wiley, Springer Nature, Elsevier, Taylor & Francis) integrate with it. Reviews are tracked and can be made public, showing you reviewed for a journal, not what you said. It's free and links to your Web of Science author profile.

ORCID reviewer record

Some journals and publishers link review activity directly to your ORCID profile. This shows up on your ORCID page alongside your publications and is visible to anyone who looks at your profile.

Journal acknowledgments

Many journals send annual "thank you" emails or list top reviewers in their January issue. These appear in PubMed and are citable on your CV. Some journals (e.g., eLife, EMBO Journal) have moved toward publishing reviews alongside papers, giving reviewers public credit.

Listing reviews on your CV

Add reviewing activity under Professional Service: "Peer reviewer for Nature Medicine, Cell Metabolism, JCI (2022–present)". You don't need to list every review. Just the journals. If you reviewed 5+ papers for one journal in a year, you can note the volume (e.g., "12 reviews for Journal of Immunology, 2024–2025").

Frequently Asked Questions

Can PhD students write peer reviews?

Yes, with their PI's knowledge and involvement. It's ethically acceptable and professionally valuable for PhD students to review manuscripts, provided their supervisor is aware, takes responsibility for the review, and ideally gives feedback before submission. Some journals have formal policies on this, so check before you start.

How long should a peer review be?

400 to 800 words is typical for a full original research article. The goal is a clear summary, numbered major concerns, numbered minor concerns, and a recommendation. Longer reviews aren't more useful. Every comment should be actionable. A three-page review full of vague concerns is less helpful than a focused 400-word review with specific, addressable points.

What should I do if I suspect fraud or data fabrication in a paper I'm reviewing?

Contact the editor directly and confidentially. Do not include your concern in the review comments visible to authors. Describe specifically what you observed (duplicate figures, implausible statistics, inconsistent data). The editor will handle it from there. COPE (the Committee on Publication Ethics) has published flowcharts for how editors should respond to suspected fraud.

References

  1. Committee on Publication Ethics (COPE). Ethical Guidelines for Peer Reviewers. 2017. Retrieved March 2026. [publicationethics.org ↗]
  2. Publons / Web of Science. Reviewer Recognition. Retrieved March 2026. [publons.com ↗]
  3. Nature Portfolio. Peer Review Policies. Retrieved March 2026. [nature.com ↗]
  4. Ware M. Peer review: recent experience and future directions. New Journal of Physics. 2011;13(1):083002. [doi:10.1088/1367-2630/13/8/083002 ↗]

Suggested Citation

APA

Manusights. (2026). How to write a peer review: A practical guide for researchers. Retrieved from https://manusights.com/resources/how-to-peer-review

MLA

Manusights. "How to Write a Peer Review: A Practical Guide for Researchers." Manusights, 2026, manusights.com/resources/how-to-peer-review.

VANCOUVER

Manusights. How to write a peer review: a practical guide for researchers [Internet]. 2026. Available from: https://manusights.com/resources/how-to-peer-review

CC BY 4.0 - share and adapt freely with attribution to Manusights (manusights.com/resources).

About these resources: Manusights is a pre-submission manuscript review service staffed by researchers with publications in Cell, Nature, Science, and related journals. These reference guides are produced as free, independent resources for the research community. No sign-up required. Data sources and methodology are cited on each page. Browse all resource guides or learn about Manusights.