What is peer review?

What reviewers actually look at

A reviewer reads a manuscript with a few questions in mind. Does the design answer the question asked, or would a different comparison group have been needed? Are the statistics suited to the data, and are the numbers in the tables consistent with the text? Do the conclusions stay within what the results show? Is anything important missing, such as a limitation, a trial registration or a reporting checklist? Most journals send a paper to two or more reviewers, and the editor reads their reports alongside the manuscript. Reviewers advise; the editor decides, and can overrule a reviewer whose objection seems unfounded.

The models, and why journals choose different ones

Elsevier describes single anonymised review, where the reviewers' names are hidden from the author, as the traditional and by far the most common model. Double anonymised review hides both sides, although Elsevier notes that reviewers can often identify the authors through writing style, subject matter or self citation. Triple anonymised review also keeps both identities from the handling editor. Open review is an umbrella term for models in which the two sides know each other, and transparent review publishes details such as the handling editor's name or the reports themselves. A journal's choice usually reflects its field: small specialties where everyone knows everyone gain less from anonymity.

The obligations on reviewers

The ICMJE recommendations set out duties that many journals adopt. Reviewers should keep a manuscript and its contents strictly confidential, must not keep a copy for personal use, and should destroy copies once the review is submitted. They should declare relationships that might bias their evaluation and step aside if a conflict exists. The recommendations also say reviewers must ask the journal's permission before using AI tools to help with a review, since such tools can produce confident text that is wrong or incomplete. Those rules exist because a reviewer sees unpublished work that a competitor would value.

Where peer review falls short

Peer review is a filter, not a guarantee. Reviewers rarely see the raw data, so they cannot catch a spreadsheet error or fabricated values that look plausible. Two reviewers can disagree completely about the same paper. Review can also be slow, which is one reason preprints spread. Conference abstracts sit somewhere lighter: a programme committee usually scores a short abstract against published criteria, often in a single round, and accepts it as a poster or a talk. That review is real, but it is not the same as full journal review of a complete paper, and a poster citing its own abstract should not describe it as peer reviewed publication.

Showing review status honestly on a board

A reader standing at a poster cannot open every reference, so the board should make review status easy to see. Cite the published version of a paper when one exists, not the preprint it started as. If a key comparison rests on a preprint, label it as a preprint in the reference itself. If the study on the board is not yet published, say so near the conclusion, for example that the manuscript is under review, rather than letting the reader assume the numbers have already survived a journal. Unpublished conference abstracts from other groups are best cited sparingly, because their full methods may never appear. Where space is tight, keep the one or two reviewed papers that anchor the argument and move the rest behind the QR code, where a curious reader can check them later.

Where it shows up in the poster builder

A poster's references are the place where review status matters most, and 34 of the 40 poster layouts carry a numbered reference block. The generator treats that block as required, so a brief that names no sources comes back with plausible looking references written to fill it, and nothing on the board marks them as invented. Every entry has to be replaced with a real source you have read before anything is printed. The same applies to numbers: key figures, chart values and flow counts come from the brief, and without them the draft still looks finished. Treat a generated draft the way a reviewer would, checking each claim against its source.

Questions people ask