Peer review: a complete guide for researchers

Peer review is the mechanism by which the scientific community evaluates, corrects, and validates scholarly work before — and increasingly after — it enters the published record. For researchers at any career stage, understanding how it actually works, not just in the abstract, is one of the most practically useful things you can do for your publishing life.

This guide covers the full picture: the different review models, the editorial workflow, how to read and respond to reviewer comments, how to conduct a review yourself, and where AI-assisted tools fit honestly and usefully into the picture.

What peer review is, and what it is not

Peer review is the evaluation of a manuscript by qualified experts who were not involved in the work. Its purpose is to catch errors, assess methodological soundness, check whether claims are supported by evidence, and advise an editor on publication decisions. It is not a guarantee of truth — the published record contains errors — and it is not a comprehensive quality score. It is one layer of a multi-layer process that includes editorial judgment, post-publication discussion, replication, and time.

Understanding this lets you engage with the process more constructively. Reviewers are not adversaries trying to block your work. They are domain experts with limited time, working under the same pressures you are, asked to give their honest technical opinion on your manuscript.

The main review models

Single-blind review is the most common model. Reviewers know the authors’ identities; authors do not know the reviewers’. Critics argue this introduces bias toward established names and institutions. Proponents say reviewer accountability is preserved.

Double-blind review anonymizes both parties. Authors must remove identifying information from their manuscript and supplementary materials. This is increasingly common in social sciences, education research, and some humanities journals.

Open peer review goes further: authors and reviewers know each other, and in some implementations the review reports are published alongside the paper. Proponents argue this increases reviewer accountability and transparency. Journals at PLOS, BMJ, and several open-access publishers now offer this model.

Post-publication review covers the evaluation that happens after a paper is published — through structured platforms like PubPeer, through commentary and letters in journals, and through the informal accumulation of citations and replication attempts. Preprint servers like arXiv and bioRxiv have accelerated this model, making it possible for the community to evaluate work before formal review begins.

How the process works, end to end

Submission and initial desk review

When you submit a manuscript, it first reaches an editor — usually a handling or associate editor rather than the editor-in-chief. Before external reviewers are ever contacted, the editor performs a desk review: does the paper fit the journal’s scope? Is it technically complete? Does it clear a basic threshold of novelty and quality?

A significant proportion of submissions — often the majority at high-selectivity journals — are rejected at this stage. This is not a commentary on the quality of your science; it is usually a scope or fit judgment. If you receive a desk rejection with a brief note about fit, take it at face value and resubmit elsewhere. Do not invest significant energy arguing it.

If the paper passes the desk review, the editor selects reviewers. This involves searching databases, consulting editorial boards, and sometimes asking authors for suggestions. Reviewing is volunteer work, and editors typically need to invite several candidates for every reviewer who accepts.

The review period

Reviewers usually have four to eight weeks, though this varies widely by field and journal. High-demand reviewers, busy periods, and reviewer attrition all slow the process. If your paper is with reviewers for longer than the stated timeline, a single polite inquiry to the editorial office is appropriate after two to four weeks past the deadline.

The editorial decision

After reviews are in, the editor synthesizes them into a decision. The most common outcomes are:

• Major revision — the paper has merit but requires substantial changes before the editor will consider it further
• Minor revision — the paper is close; targeted changes are needed
• Reject with invitation to resubmit — not the same as an outright rejection; the editor sees a path forward if fundamental issues are addressed
• Accept — relatively rare without at least minor revision
• Reject — the paper is not suitable for this journal in its current form, or the issues are too fundamental to fix in revision

The editor’s letter usually summarizes the reviewers’ concerns and often indicates which points the editor considers most important. Read this letter first. Editors sometimes make decisions that cut across what the reviewers said, and those editorial signals carry more weight than individual reviewer preferences.

How to read reviewer comments

This is where many researchers lose time and energy. Reading reviewer comments immediately after receipt, when the emotional response is strongest, is rarely productive. Put the decision letter aside for twenty-four to forty-eight hours if the timeline allows.

When you return to it, read everything before you start reacting to any of it. Reviewer comments often form a coherent picture once you can see the whole set. What looks like a devastating critique in isolation may be a reasonable request when you understand the broader concern.

Sort comments into categories:

• Genuine errors — misinterpretations, methodological weaknesses, or unsupported claims that require real changes
• Legitimate requests for clarity — the reviewer was confused; the paper needs to do a better job explaining something
• Requests for additional experiments or analysis — evaluate these against scope, feasibility, and what the editor’s letter says
• Disagreements — the reviewer holds a different view; you can engage with these respectfully and explain your reasoning
• Requests you cannot fulfill — be honest and explain why

The key discipline is to separate the request from the phrasing. Reviewers vary widely in how diplomatically they write. Some are terse; some are pointed. The substance usually includes something addressable.

Writing the response-to-reviewers letter

Your response letter is as important as your revised manuscript. Editors read it first. A well-organized, respectful, thorough response letter signals that you have taken the review seriously and will make reviewing the revision much easier.

Structure: open with a brief paragraph thanking the editors and reviewers and summarizing the major changes. Then address every comment, numbered to match the original review, with two parts: (1) a prose response explaining what you did or why you disagree, and (2) the specific text you added or changed, shown verbatim.

Here is an example of the format for a single comment:

Reviewer 2, Comment 3: “The authors claim their method outperforms the baseline, but they do not provide confidence intervals for the primary metric. This makes it impossible to assess whether the difference is meaningful.”

Response: We thank the reviewer for this point. They are correct that confidence intervals were not reported. We have now computed bootstrapped 95% confidence intervals for all primary comparisons and added them to Table 2 and Figure 3. The revised text reads:

“Mean accuracy was 84.3% (95% CI: 82.1–86.5%) for our approach versus 79.6% (95% CI: 77.4–81.8%) for the baseline, a difference that is both statistically reliable and practically meaningful given the clinical context (see §3.2).”

For a fuller worked example and templates, see our guide to response to reviewers samples.

If you disagree with a reviewer, say so — respectfully and with evidence. Editors respect authors who defend their methodological choices with good arguments. What editors do not respect is authors who ignore a comment or respond with vague assurances that the concern has been addressed.

How to peer review a paper yourself

If you have been asked to review a manuscript, you are being asked to provide a service to your field. Review well, and you improve the literature. Review poorly, and you waste everyone’s time.

Before you accept

Decide whether you genuinely have the expertise to evaluate the paper. It is better to decline than to submit a review outside your competence. Also check for conflicts of interest: are you a collaborator, a competitor with an active competing paper, or in another position that could bias your judgment? Declare these to the editor rather than trying to assess them unilaterally.

Reading the manuscript

Read the paper twice. The first read is for overall impressions: what is the question, what did they do, what did they find, is the conclusion supported? The second read is for detail: methods, statistics, data presentation, claims against evidence.

Take structured notes under the headings that matter most: significance, methodological soundness, clarity, limitations, and presentation.

What a good review covers

A thorough review addresses:

• Significance: Is the question worth asking? Does the paper advance understanding in a meaningful way?
• Methods: Are the study design and analysis appropriate for the question? Are sample sizes justified? Is the statistical approach sound?
• Results: Are figures and tables accurate and interpretable? Are results reported with appropriate uncertainty?
• Interpretation: Do the conclusions follow from the data? Are alternative explanations adequately considered?
• Prior literature: Is the paper appropriately positioned relative to existing work? Are there important omissions?
• Reproducibility: Is there enough methodological detail for someone to replicate the work?

For detailed checklists and discipline-specific considerations, the guide on how to peer review a paper covers the full process step by step.

Writing the review

Write your review in two parts: a summary paragraph for the editor (confidential; can include your recommendation and a frank overall assessment) and the comments to the authors (your technical critique, which they will see). Be specific. “The methods section is unclear” is not helpful. “The authors do not specify the software version or random seed used for the clustering algorithm, which limits reproducibility” is.

Common reasons papers are rejected, and how to pre-empt them

Insufficient novelty

The contribution needs to be clear before you submit, not as an afterthought in the discussion. Establish what was not known, what you found, and why that matters — in the abstract and again in the introduction. If you cannot articulate the specific gap you are addressing in one or two sentences, the paper is not ready.

Methodological weaknesses

Reviewers notice underpowered studies, inappropriate statistical tests, missing controls, and convenience samples presented as representative samples. These are best caught in manuscript preparation, not in revision. Having a statistician review your analysis before submission is almost always worth it.

Misalignment with journal scope

Submitting a narrow technical paper to a broad-audience journal, or vice versa, is a common and avoidable error. Read the aims and scope carefully. Look at recent publications. If the journal publishes review articles and you have submitted an empirical study outside the editorial sweet spot, expect a desk rejection.

Overclaiming

Language that overstates what the data support — “proves,” “demonstrates definitively,” “the first study ever to show” — invites skeptical review and signals that the authors have not engaged critically with their own limitations. Calibrated language is a marker of scientific maturity.

Poor presentation

Reviewers are human, and manuscripts that are difficult to read accumulate doubts. Clear writing, well-designed figures, and logical structure do not substitute for good science, but they do affect how your science is received. Proofreading and editing are not vanity — they are part of communicating research responsibly.

Where AI-assisted review fits

AI tools have entered manuscript preparation in a way that is neither entirely new nor entirely uncontroversial. The honest picture: AI-assisted review can be genuinely useful for certain tasks; it cannot replace expert human judgment on what matters most.

What AI does well in this context is structural and systematic. A well-designed AI peer review tool can check internal consistency — whether claims in your abstract match results in your tables, whether figures are discussed accurately in the text, whether your limitations section addresses the obvious objections your methods raise. These are tasks that benefit from systematic, dispassionate attention to the full document, which is difficult to maintain after you have been immersed in the work for months.

What AI does not do is evaluate scientific significance, judge whether a method is appropriate for your specific domain’s norms, or catch subtle conceptual errors that require deep subject-matter expertise. Those judgments still belong to qualified human reviewers.

PerfectPaper is built around this honest division of labor. It functions as a research paper reviewer for your draft — giving you referee-grade feedback on methods, evidence, logic, and internal consistency before you submit. The goal is to surface the issues that reviewers are likely to raise so that you can address them on your own schedule, not under revision pressure. It is not a substitute for human review; it is preparation for it.

If you want to see what that kind of feedback looks like on your own manuscript, you can start a free review and upload a paper today.

For a broader look at how AI peer review is changing manuscript preparation, and what to look for when evaluating peer review software, those guides cover the landscape in detail.

Frequently asked questions

What is the difference between single-blind and double-blind peer review?

In single-blind review, reviewers know who the authors are, but authors do not know the reviewers’ identities. In double-blind review, both parties are anonymous to each other. Double-blind review is designed to reduce bias based on author reputation or institutional affiliation, though it is not a perfect solution since authors can sometimes be identified from their methods, data, or writing style.

How long does peer review typically take?

It varies considerably by field and journal. In many life sciences and biomedical fields, review cycles of three to six months are common for initial decisions, with additional time for revisions. In some humanities disciplines, review can take considerably longer. Computational and engineering fields have seen shorter cycles emerge in part through conference publishing models. If you have not received a decision well past the journal’s stated timeline, a polite inquiry to the editorial office is appropriate.

Can I suggest reviewers when I submit?

Most journals offer this option. Suggested reviewers are not guaranteed to be used, but editors often find them helpful — particularly for highly specialized work where the pool of qualified reviewers is small. The suggested reviewers should be genuine experts with no conflicts of interest. Do not suggest collaborators or close colleagues; editors can recognize these patterns and it reflects poorly on the submission.

What should I do if I think a reviewer was wrong?

Address the comment directly and respectfully in your response letter. Explain your reasoning, cite evidence, and be specific. If the reviewer’s concern reflects a genuine ambiguity in the manuscript, clarify the manuscript even if you believe your original version was defensible. Editors generally respect authors who engage with critiques substantively rather than accepting all changes uncritically. What editors do not find useful is authors who dismiss reviewer concerns without explanation.

Is it acceptable to use AI tools when preparing a manuscript or response letter?

Norms are evolving and vary by journal, so check the specific journal’s policy. Many journals now require disclosure of AI assistance in writing. The more important practical distinction is between using AI to assist your thinking and writing versus using AI to generate content you present as your own unassisted work. Using an AI peer reviewer to get structured feedback on your manuscript before submission is a preparation activity, analogous to asking a colleague to read a draft; it does not raise the same disclosure questions as using AI to write your results section.

How do I know if a journal uses rigorous peer review?

Look for journals indexed in major disciplinary databases, with identifiable editorial boards whose members you can verify as active researchers in the field, clear statements of their review process, and published turnaround statistics. Membership in organizations like the Committee on Publication Ethics (COPE) and use of DOI registration are positive signals. If a journal’s acceptance rates seem implausibly high or its turnaround implausibly fast, investigate before submitting.

Peer review is imperfect, slow, and occasionally frustrating. It is also the most durable mechanism the research community has for collective quality assessment. Learning to engage with it skillfully — as an author, as a reviewer, and eventually as an editor — is a professional capability that compounds over a career. The researchers who take it seriously, on all sides, are the ones who end up with the clearest signal in the literature.

If you are preparing a manuscript and want a structured read before it goes out, start a free review with PerfectPaper.