WHAT IS PEER REVIEW? A COMPLETE GUIDE.

By María García-Puente · Editorial · Updated April 2026

Peer review is the process by which a manuscript submitted to a scholarly journal is evaluated by independent experts in the same field, usually two or three, before it's published. It serves three purposes: filtering out work that doesn't meet basic standards, improving the methodological quality of what does, and giving editors outside judgment to inform a decision that would otherwise rest with a single person.

It's worth saying upfront what peer review is not, because that's where most misunderstandings start. It's no guarantee that a paper is true, or that its results will replicate. It's poor at catching deliberate fraud, since it assumes good faith on the part of whoever signs the manuscript, and it almost always takes longer than anyone would like. Even with those limits, it remains the mechanism that best sustains trust in the system: it doesn't certify that a piece of work is good, but it does confirm that someone who knows the field has read it critically before it reaches the reader.

IN ONE SENTENCE

Peer review doesn't guarantee that a paper is good; it guarantees that it was read critically before publication.

1. Submission. The author uploads the manuscript through the journal's platform and confirms it meets the guidelines: formatting, length limits, conflict-of-interest disclosure, and ethical requirements.
2. Editorial screening (desk review). Before troubling any reviewer, the editor reads the work and decides whether it fits the journal and has enough substance to go out for review. A large share of submissions are rejected at this stage, sometimes within days, almost always because they don't fit the journal's scope rather than because of poor quality.
3. Reviewer assignment. If it clears the screening, the editor looks for two or three experts who know the topic and have no conflict with the authors. Finding them, and getting them to say yes, is in practice the slowest phase of all.
4. Review. Reviewers read the manuscript and return a report with comments and a recommendation: accept, request revisions, or reject. They don't decide; they advise.
5. Editorial decision. With the reports in hand, the editor makes the decision and communicates it to the author. If changes are requested, a new round opens, which can repeat several times.
6. Production and publication. Once accepted, the manuscript moves to copyediting, layout, and publication, with its DOI assigned.

• Single-blind: the reviewer knows the author's identity, but the author doesn't know who reviewed the work. It's the most common model and protects reviewers from retaliation, though it leaves the door open for the author's reputation to influence the judgment.
• Double-blind: neither the author nor the reviewer knows the other's identity. The idea is to reduce biases around reputation, gender, or country of origin. It works better on paper than in practice, because in small fields reviewers often guess who wrote the paper from the topic, the citations, or the writing style — so the anonymity is more fragile than it looks.
• Open: author and reviewer know each other, and some journals also publish the review reports alongside the article, sometimes signed. It gains in accountability and transparency what it loses in comfort, because no one writes the same way knowing their report will carry their name.
• Post-publication: the article is published first, and the community comments on and evaluates it afterward, as on PubPeer. It doesn't replace pre-publication review; it complements it, recognizing that critical reading of a paper doesn't end the day it comes out.

A good editor isn't the one who rejects the most; it's the one who gives clear reasons for rejection and publishes with consistency.

• Accepted. The least common outcome on a first round. It means the work will be published, sometimes with minor stylistic tweaks.
• Minor revision. The substance is convincing and only details need adjusting: clarifying a method, fixing references, qualifying a conclusion. It usually resolves in a single round.
• Major revision. There's interest, but also substantive doubts: analyses to redo, arguments to strengthen, data to complete. It's not a no; it's a "not yet, and here's how it could become a yes."
• Rejection. This can be due to a poor fit with the journal, serious methodological problems, or insufficient contribution. A rejection with good comments is valuable information for resubmitting the work to a better-suited journal.

The system fails when deadlines drag on forever, when reviewers simply can't find the time, when some authors look for shortcuts, and when a journal simply stops reading what it receives with real attention. All of this happens, and all of it is managed with clear processes and sustained human judgment.

In recent years, artificial intelligence has joined that classic picture, and it now runs through both sides of the process: authors drafting with language models, and reviewers tempted to delegate their report to one. Bodies like the ICMJE and publishers themselves have started setting rules, and the emerging consensus is sensible: AI can help with drafting or spotting problems, but responsibility for judgment — and confidentiality of the manuscript — remain human. This is a space that's still being written, and it's worth following closely.

If you've read this far because you manage a journal, or you're thinking about tightening up its editorial workflow, most of the problems this guide describes aren't solved with more goodwill, but with process: evaluation templates that guide reviewers, reminders that keep a round from stalling, deadlines that are communicated and kept, and editorial judgment sustained over time. That's exactly what we do in our editorial management work, where we help scholarly journals keep peer review on schedule and rigorous. And if your interest runs more toward evidence synthesis, our piece on systematic reviews and PRISMA might be useful.

If this sounds like your journal, let's talk.