Peer review is the vetting step journals use before publishing a scientific paper: an editor screens the submission, then sends it to independent experts who judge its methods, evidence, and conclusions before a decision to accept, revise, or reject is made. That is how Nature Portfolio journals describe the process in the editorial policy they publish for authors and reviewers, and the broad outline holds across most research journals, though the details vary by field and publisher.
What actually happens after a paper is submitted?
A manuscript first lands with an editor, not a reviewer. Editorial staff read every submission and reject, "promptly without external review," papers that are out of scope, poorly reported, or clearly not competitive, according to Nature Portfolio's editorial policy. Only papers judged to have a realistic shot at publication move on to formal review.
For those that advance, an editor typically lines up two or three outside reviewers, "sometimes more if special advice is needed." Each reviewer works alone, without seeing the others' comments, and reports back on a timeline the editor sets. When reports come in, the editor does not simply tally votes for or against; the policy describes editors weighing "the strength of the arguments raised" before choosing among four outcomes: accept, request revisions, reject with an invitation to resubmit, or reject outright.
Who reviews the work, and what are they looking for?
Reviewers are researchers working in the same or an adjacent field, chosen by the editor for relevant expertise rather than random assignment. Per Nature Portfolio's policy, they are asked to weigh in on a specific list: the key results and what is novel about them, the validity of the methods and any flaws, whether the data quality and statistical approach hold up, whether the conclusions are actually supported by the evidence presented, and whether the references and presentation are adequate.
That list matters because it shows what peer review is checking for and what it is not. Reviewers are evaluating internal consistency — does the evidence in this paper support what the authors claim it supports — not independently repeating the experiment. Confidentiality is expected throughout: reviewers agree not to discuss or use an unpublished manuscript before it appears in print.
Single-blind, double-anonymized, or open — why does the format matter?
Journals run peer review under different rules about who knows who. In traditional single-blind review, the reviewer's identity stays hidden from the authors unless the reviewer chooses to sign their report, per Nature Portfolio's policy. Some journals instead offer double-anonymized review, in which authors' identities are also withheld from reviewers for the full evaluation — an option meant to reduce the chance that a reviewer's judgment is shaped by an author's reputation or institution rather than the work itself.
A third and newer model, transparent peer review, publishes the reviewer reports, the authors' responses, and the editor's decision letters alongside the accepted paper. That model trades reviewer anonymity for a public record of exactly what was questioned and how the authors answered — something a reader normally never sees.
One newer wrinkle: reviewers are barred from uploading a manuscript under review to a generative AI tool, and must disclose it if AI assistance played any role in their evaluation, according to Nature Portfolio's current policy.
How is reviewing a grant proposal different from reviewing a journal paper?
Peer review is not limited to papers already written up. Funding agencies use a parallel process, before a study is even run, to decide which proposals get money. The National Science Foundation calls its version merit review: program officers identify outside experts, who serve as individual reviewers, a panel, or a mix of both, according to the agency's official description of the process.
NSF judges every proposal against two board-approved criteria: intellectual merit, meaning the work's potential "to advance knowledge and understanding within its own field or across different fields," and broader impacts, meaning its potential to "benefit society and contribute to the achievement of specific, desired societal outcomes." Reviewer input then feeds into a funding decision that also weighs program priorities, such as whether NSF's overall portfolio spans a diverse mix of approaches. The agency states it aims to notify applicants of a decision within six months.
The distinction matters for how readers should treat each kind of claim. A funded grant means reviewers judged a proposed study worth trying — it says nothing yet about what the study will find. A peer-reviewed paper means reviewers judged the finished analysis defensible enough to publish. Neither stamp certifies that the underlying result is correct.
What does peer review not guarantee, and who checks the checkers?
Peer review is a filter applied before publication, run by reviewers who — like the authors — can miss errors, share blind spots, or simply disagree. It is not a replication of the experiment and not a guarantee that a published finding will hold up. That is part of why science treats a single paper, however well reviewed, as provisional rather than final: later studies, replication attempts, and post-publication scrutiny are what let a result move from "one team's peer-reviewed finding" toward broader confidence.
When something does go wrong after publication — a flawed method surfaces, data cannot be reproduced, an image turns out to be duplicated — journals rely on a separate track: corrections, expressions of concern, or retractions, typically after an internal investigation. The Committee on Publication Ethics, an international membership body for journal editors, exists specifically to give editors a shared forum for weighing those post-publication integrity problems, according to a description of the group published by the National Library of Medicine. It does not review individual manuscripts itself; it sets shared expectations for how editors should handle disputes once a paper is already in print.
None of this means peer review is decorative. It is the reason a claim that reaches a journal has already been read skeptically by people qualified to catch an obvious flaw before the public sees it. It just is not, and was never designed to be, the last word on whether a finding is true.
For a related briefings perspective, read How does an mRNA vaccine actually work?.