How peer review works, and what a preprint is

A preprint is a complete research paper that its authors post online before a journal has reviewed it. Peer review is the journal's check: an editor sends the paper to outside experts, weighs their reports and decides whether to publish it, ask for changes or turn it down. A peer-reviewed article has been through that check. A preprint has not, or not yet. Both can be useful, but they carry different weight.

How journal peer review works

Journals set their own rules, so it helps to look at one in detail. PLOS One, which takes research from across the natural sciences, medicine, engineering and related social sciences, publishes its process step by step:

1. Journal staff run initial checks, covering competing interests, compliance with editorial policies and ethical standards, financial disclosures and data availability. Papers can be sent back to the authors at this stage.
2. An academic editor with relevant expertise reads the manuscript against the journal's criteria and decides whether reviews from outside experts are needed.
3. The editor invites reviewers, chosen for their expertise, publication history and past reviews. Once they agree, reviewers typically have 10 days to submit a report.
4. The editor weighs the reports and their own reading, and makes one of four decisions: reject, major revision, minor revision or accept.
5. A revised paper usually returns to the same editor, who may decide alone or send it back to reviewers.

PLOS One uses single-anonymized review, in which reviewers stay anonymous unless they choose to sign their reports. Authors whose papers are accepted can opt to publish the full peer review history alongside the article, including the editor's decision letters, the reviews and the authors' responses. The journal says its editors decide based on scientific rigor, “regardless of novelty.”

What a preprint server does, and does not do

Preprint servers post papers quickly, after a light screen rather than a review. bioRxiv, for the life sciences, and medRxiv, for the health sciences, are both run by the nonprofit openRxiv. Both say their articles “are not peer-reviewed, edited, or typeset before being posted online.” Every submission is screened for offensive or nonscientific content, for material that might pose a health risk or be considered dual-use research of concern, and for plagiarism. Both servers state that posting a paper implies no endorsement of its methods, assumptions, conclusions or scientific quality.

medRxiv adds a direct warning about health research. Its preprints, it says, “should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.”

arXiv, which describes itself as a forum providing “rapid distribution of new research,” has moderators check submissions, mostly before they go public. They can reclassify or decline papers, for example for lacking originality or containing plagiarized material. But arXiv states plainly that its moderation “is not a peer-review process.” It also asks authors to report significant use of generative AI tools and holds every listed author responsible for the full content.

How a preprint becomes a paper

The two routes often meet. bioRxiv says most research journals allow authors to post a preprint before publication, and authors can submit a revised version at any time until a journal formally accepts the paper. Revisions keep the same DOI, the digital object identifier used to cite the paper. bioRxiv usually adds a link to the published version within several weeks of journal publication, and it can send a manuscript's files directly to participating journals.

Once posted, a preprint normally stays up. bioRxiv and medRxiv both say posted articles are citable and cannot be removed, although bioRxiv reserves the right to remove plagiarized work or work that breaks ethical standards. arXiv says an announced paper becomes part of the permanent scholarly record, though it can withdraw papers that break its policies.

How to read each kind of study

• Check where it appeared. This week's Mars dust study, for example, names its journal, the Proceedings of the National Academy of Sciences, and carries a DOI.
• For a preprint, treat the findings as preliminary, especially on health. medRxiv's own advice is not to rely on them for clinical decisions.
• Look for later versions. A preprint may have been revised, or linked to a published version, since it first appeared.
• Look for the review history. PLOS One, for one, lets authors publish the reviewers' reports and their responses alongside the article.

Our report on the Perseverance dust study shows a journal paper explained with its limits, including what the measurements do not show.