How to Review a Journal Paper: Step-by-Step Guide
An editor has asked you to review a manuscript and nobody ever taught you how. This guide explains how to review a journal paper from the invitation email to the final recommendation, with a section-by-section checklist and a report template you can reuse.
What the editor wants from you
A review answers two questions. For the editor: is this work sound, new, and right for this journal? For the authors: what would make it better? You are an adviser, not the judge. The editor makes the decision, usually after reading two or three reports, so your job is to give clear reasons rather than a verdict.
Underneath that sits a set of professional duties: confidentiality, fairness, and meeting the deadline. We cover those in a reviewer's main responsibility.
Step 1: Decide whether to accept the invitation
The invitation normally includes the title and abstract. Before you click "accept", check four things.
- Expertise. Can you judge the core of this paper? You do not need to cover every technique, but you must be able to assess the main claim.
- Conflict of interest. Have you worked with the authors recently, do you share an institution, or are you competing on the same question? If so, tell the editor and let them decide.
- Time. A careful first review takes several hours spread over a few days. If you cannot do it by the deadline, say no now.
- Journal instructions. Read the reviewer guidelines. Some journals use a scoring form, some want the recommendation only in confidential comments, and policies on AI tools differ.
Declining is fine and carries no penalty. Reply quickly either way, and suggest another reviewer if you can.
Step 2: Read the paper once for the big picture
Read from start to finish without stopping to fix details. Wiley's reviewer guide recommends this skim to form a first impression and to find any major flaw early. Keep notes with page and line numbers. By the end you should be able to answer:
- What question does the paper ask, and is it worth asking?
- Is the approach capable of answering it?
- Do the conclusions follow from the evidence shown?
- Is it new, or does it repeat what is already known?
- Does it fit this journal's readers?
Then write two short paragraphs: one that summarizes what the authors did and found, and one that says what the paper adds to the field. These become the opening of your report. If you find you cannot summarize the paper, that is itself a finding to report.
Even when you spot a serious flaw, finish reading. You need the full context to be sure the flaw is real, and the authors deserve to hear what was good.
Step 3: Read it again, section by section
The second pass is slower. Keep the figures and tables visible beside the text. Use this checklist, drawn from the guidance published by Springer Nature and by DeHart and colleagues in the HCA Healthcare Journal of Medicine.
| Section | Questions to ask |
|---|---|
| Title and abstract | Does the title say what the study is about? Does the abstract match the paper, with nothing in it that is missing from the main text? Could a researcher outside the specialty follow it? |
| Introduction | Is the literature current and fairly represented? Is the gap clear? Is the research question or hypothesis stated, and does the background justify it? |
| Methods | Could someone repeat the study from this description? Are the sample, inclusion and exclusion criteria, controls, and outcome measures defined? Is ethics approval or consent reported where needed? Are there sources of bias the design does not handle? |
| Statistics | Is the sample size justified? Do the tests suit the data? Are repeated measurements treated as independent when they are not? Are exact p-values and effect sizes given, with measures of uncertainty? |
| Results | Are all results traceable to a method? Are they stated plainly, without interpretation mixed in? Is important data missing or marked "not shown"? |
| Figures and tables | Can each be understood from its caption alone? Are axes scaled honestly? Does anything look edited or duplicated? Is the same data repeated in text, table, and figure? |
| Discussion and conclusions | Do the conclusions stay within what the data support? Are contrary findings cited? Are limitations stated? Do the conclusions answer the aims set out in the introduction? |
| References | Are key studies missing? Is the list balanced, or heavy with the authors' own work? Do the sources cited for central claims say what the authors claim? |
The methods section deserves most of your attention. Weak writing can be fixed in revision; a design that cannot answer the question usually cannot.
If part of the paper is beyond your expertise, most often the statistics, tell the editor so they can ask a specialist. Do not guess.
Step 4: Sort your notes into major and minor issues
Editors and authors need to know which comments decide the paper's fate and which are polish.
| Major issues | Minor issues |
|---|---|
| Conclusions the data do not support | Ambiguous sentences or undefined terms |
| Flawed or outdated method; missing controls | Missing or incorrect references |
| Too little data, or data that contradict each other | Unit, number, or labelling errors |
| Similar published work that the authors ignore | Figures that could be clearer or combined |
| Ethical concerns | Typos that change meaning |
You are not the copy editor. Mention language only where it gets in the way of understanding, and say so in one general comment rather than listing every error.
Ask for extra experiments or analyses only when the paper cannot stand without them. Almost any study could be extended; that is not a reason to hold it up.
Step 5: Write the review report
Most reports follow the same shape. Number your points so the authors can answer each one.
Review report template
- Summary (one or two paragraphs). What the paper sets out to do, how, and what it found. What it contributes. Its main strengths. Your overall view in a sentence.
- Major issues (numbered). For each: what the problem is, where it is (page and line), why it matters, and what would resolve it.
- Minor issues (numbered). Short, specific, with locations.
- Confidential comments to the editor (if the journal offers them). Suspected misconduct, your own limits as a reviewer, and your recommendation if the journal wants it kept out of the comments to authors.
A report does not need to be long. A page that identifies the two things that matter is more useful than five pages of line edits.
Tone: write what you would say in person
Criticize the work, never the authors, and make every criticism something they can act on. DeHart and colleagues give a useful pair of examples: "The authors' conclusions are absurd" insults and helps nobody, while asking the authors to check that their conclusions reflect their findings tells them what to do. They also suggest writing in the first person ("I had difficulty matching the results to the final paragraph") because it reads as a colleague's comment rather than a ruling.
Physicist Steven Thomson, in his advice for first-time reviewers, adds a practical habit: draft the review, leave it for a few days, and reread it before sending. A second look often shows that you misread something, or that you missed a problem.
Step 6: Make your recommendation
| Recommendation | Use it when |
|---|---|
| Accept | The work is sound and clear as it stands. This is rare for a first submission. |
| Minor revision | The science is sound and the remaining changes are small enough for the editor to check without sending it back to you. |
| Major revision | The work is promising but has significant problems in analysis, interpretation, or presentation that you believe the authors can fix. You will probably be asked to look at the revised version. |
| Reject | The flaws cannot be fixed by revision, the work is not original, or it does not fit the journal. |
If you recommend rejection, still explain why and what the authors could do differently. They will probably submit elsewhere, and a clear report improves the paper wherever it ends up. Make sure the comments the authors see match what you tell the editor: a polite report paired with a harsh confidential note leaves authors confused about why they were rejected.
When a revised manuscript comes back, check the authors' responses against your original points, and be ready to accept that you were wrong on some of them.
How to review articles for a journal that are not original research
Not everything you are sent will be an experimental paper. The steps above still apply, but the weight shifts.
- Review articles. There may be no methods to check. Judge whether the review is timely, whether it covers the literature broadly and accurately, whether the choice of studies is balanced, and whether the structure helps the reader. Ask whether it points to useful directions for future work.
- Systematic reviews and meta-analyses. Here the method is everything: the search strategy, inclusion criteria, quality assessment of included studies, and how results were combined.
- Qualitative studies. Look for a recognized method that suits the question, a clear account of how data were collected and analyzed, and enough quoted material to show the findings go beyond the authors' opinions.
- Case reports and short communications. Ask whether the case teaches something new and whether the claims are kept in proportion to a single observation.
Common mistakes first-time reviewers make
- Listing every flaw without saying which ones matter.
- Rewriting the paper into the one they would have written.
- Spending the review on grammar.
- Asking for citations to their own work without good reason.
- Giving a verdict with no reasons, such as "not suitable for this journal".
- Sharing the manuscript with a colleague, or pasting it into an AI tool, without checking the journal's rules.
- Sitting on the review past the deadline without telling the editor.
Confidentiality and ethics checks
Treat the manuscript as confidential before, during, and after the review. Do not use its ideas or data. If you suspect plagiarism, image manipulation, duplicate publication, or fabricated data, describe what you saw in the confidential comments to the editor and let the journal investigate. Do not contact the authors.
Frequently asked questions
How long does it take to review a journal paper?
It depends on the length of the paper and how close it is to your expertise. Plan for several hours spread over a few days, including time to reread your draft before sending it.
How long should a peer review report be?
Long enough to summarize the paper and explain each major issue. One to two pages is common. A short report that identifies the key problems is more useful than a long list of line edits.
Should I correct grammar and spelling when reviewing a paper?
Only where language prevents you from understanding the work. Note the problem in one general comment. Copy editing is not the reviewer's job.
Can I decline an invitation to review?
Yes. Decline if the paper is outside your expertise, you have a conflict of interest, or you cannot meet the deadline. Reply quickly and suggest another reviewer if you can.
Can I use AI tools to help review a journal article?
Check the journal's policy first. Many journals do not allow reviewers to upload a manuscript or any part of it to a generative AI tool, because it breaches confidentiality.
What is the difference between major and minor revision?
Major revision means the paper has significant problems that the authors can probably fix, and it will usually be reviewed again. Minor revision means the science is sound and only small changes are needed.