How to Write a High-Quality Peer Review

A practical guide for reviewers committed to rigorous, constructive, and ethical manuscript evaluation — across all disciplines.

Peer review is the cornerstone of scholarly publishing. It serves as the primary mechanism through which the academic community validates research before it enters the public record. Yet despite its importance, most researchers receive little formal training in how to conduct a review. The result is a wide disparity in review quality — from reports that genuinely help authors strengthen their work, to those that are superficial, unhelpful, or even counterproductive.

This guide outlines a structured approach to writing a peer review that meets the standards expected by high-quality international journals. Whether you are reviewing a manuscript in the natural sciences, social sciences, engineering, humanities, or any interdisciplinary domain, the principles described here apply universally.

Before You Begin: Deciding Whether to Accept

When you receive a review invitation, your first task is to determine whether you are the right person for the job. Consider three questions: Do you have sufficient expertise in the subject matter to provide a meaningful evaluation? Can you complete the review within the journal’s requested timeframe? Are you free of any conflicts of interest with the authors, their institutions, or the research itself?

If the answer to any of these is no, decline promptly. A timely decline is far more valuable to the editorial process than a delayed or superficial review. If you have partial expertise — for example, you can evaluate the methodology but not the domain-specific literature — communicate this to the editor. In a multidisciplinary journal, this kind of transparency is especially important, as manuscripts often span fields that no single reviewer can fully cover.

Step 1: Read the Manuscript in Full — Twice

Begin with a complete read-through without making notes. The purpose is to understand the manuscript’s overall argument, its structure, and the authors’ main claims. On your second reading, engage critically. Annotate passages that raise questions, mark statistical claims that seem unsupported, flag sections where the logic breaks down, and note where additional context would strengthen the argument.

Pay particular attention to the abstract and conclusions. These are the sections most widely read, and they must accurately represent the work described in the manuscript. Overstated conclusions — those that go beyond what the data actually support — are among the most common and consequential problems in submitted manuscripts.

Step 2: Structure Your Report

A well-organized review is easier for both the editor and the authors to act upon. The recommended structure, used by the majority of high-quality journals, follows a clear hierarchy.

Summary paragraph

Open your review with a brief summary — in your own words — of what the manuscript reports and what it claims to contribute. This demonstrates that you have read the work carefully and helps the editor understand your interpretation, which may differ from that of other reviewers. Keep it to three or four sentences.

Overall assessment

State your general evaluation of the manuscript’s strengths and weaknesses. Is the research question significant? Is the approach sound? Does the manuscript make a meaningful contribution to the field? Be honest but measured. If the work has fundamental problems, say so clearly; if it has merit but needs improvement, acknowledge both.

Major concerns

These are issues that, if unaddressed, would prevent the manuscript from being publishable. Examples include flawed experimental design, inappropriate statistical methods, conclusions not supported by the data, missing controls, or a failure to engage with directly relevant prior work. Be specific: cite the relevant section, figure, or table number. Explain why the issue matters and, where possible, suggest how it might be resolved.

Minor concerns

These are issues that should be addressed but would not, on their own, prevent publication. Missing references, unclear figure labels, minor inconsistencies in terminology, or suggestions for improved data presentation belong here. Again, reference specific locations in the manuscript.

Step 3: Evaluate the Core Dimensions

Regardless of discipline, every manuscript should be evaluated along several fundamental dimensions.

• Originality. Does this work present new findings, a novel method, or a fresh perspective? Has similar work been published before? If so, does this manuscript add something substantive?
• Methodology. Is the research design appropriate for the question being asked? Are the methods described in sufficient detail to allow replication? Are limitations acknowledged?
• Data quality. Are the data presented clearly and accurately? Are statistical analyses appropriate, and are uncertainties properly reported? Are figures and tables well-constructed and necessary?
• Interpretation. Do the conclusions follow logically from the results? Are alternative explanations considered? Are claims proportionate to the evidence?
• Contextualization. Is the work situated within the relevant literature? Are key references cited? Does the discussion provide sufficient context for readers who may not be specialists in the subfield?
• Clarity. Is the manuscript well-written and logically organized? Are technical terms defined where appropriate? Note that reviewers are not expected to provide line-by-line copyediting — focus on clarity of argumentation rather than grammar.

Step 4: Write for the Right Audience

Your review has two audiences: the editor and the authors. The editor needs your assessment to make an informed decision. The authors need your feedback to improve their work. Write with both in mind.

Be constructive. A review that simply lists problems without suggesting paths forward is far less useful than one that identifies issues and offers concrete guidance. At the same time, do not soften genuine concerns — editors depend on honest assessments, and authors benefit more from frank, respectful feedback than from vague encouragement.

A thoughtful, well-presented evaluation of a manuscript, with tangible suggestions for improvement and a recommendation supported by the comments, is the most valuable contribution a reviewer can make. — ACS Reviewer Lab

Step 5: Provide Your Recommendation

Most journals ask reviewers to select from a standard set of recommendations: accept, minor revision, major revision, or reject. Your recommendation should be consistent with the tone and substance of your written report. If your comments are largely positive, a recommendation to reject will confuse both the editor and the authors. Conversely, if you have identified fundamental flaws, a recommendation for minor revision sends a contradictory signal.

If you are uncertain, it is entirely appropriate to state that your recommendation depends on how the authors address specific major concerns. The editor will weigh your assessment alongside those of other reviewers.

Confidential Comments to the Editor

Most review systems provide a separate field for confidential comments visible only to the editor. Use this space to flag any ethical concerns — suspected plagiarism, data fabrication, undeclared conflicts of interest, or substantial overlap with other published work. This is also where you can note limitations in your own expertise and suggest that additional specialist input may be needed.

Do not use confidential comments to make critical remarks about the manuscript that you are unwilling to share with the authors. If a concern is substantive enough to influence the editorial decision, the authors have a right to respond to it.

A Note on Ethics and AI

Peer review is a confidential process. The manuscript you receive is an unpublished work entrusted to your professional judgment. Do not share it with colleagues without the editor’s permission, do not use the findings to advance your own research before publication, and do not upload the manuscript to AI tools or other third-party platforms, as this may compromise confidentiality and copyright.

If you choose to use generative AI tools to help refine the language of your review report, this must be disclosed to the editor, consistent with current guidance from the Committee on Publication Ethics (COPE) and the journal’s own policies. The intellectual content and judgments in the review remain your responsibility.

— — —

This article reflects guidelines aligned with COPE Ethical Guidelines for Peer Reviewers and best practices. Research and Science Today is committed to maintaining the highest standards of peer review integrity across all disciplines.