A link appears in a group chat or a news story: researchers have posted a new study on arXiv, bioRxiv, or medRxiv. The PDF looks like a paper. It has a title, a methods section, figures, and a date. It may already be written up as if the finding were established. For a reader who is not inside the field, the most useful first question is not "Is this true?" It is "What kind of document is this?"
The core idea is simple: a preprint is a manuscript that the authors have posted publicly before formal peer review. The posting makes the work citable and time-stamped. It does not, by itself, certify that the results have been checked and endorsed by independent reviewers.
That practice is older than the current platforms, but the modern, centralized version has a clear starting point. On 14 August 1991, the physicist Paul Ginsparg launched an automated repository for electronic preprints at Los Alamos National Laboratory, originally as an email archive for high-energy theory papers at hep-th@xxx.lanl.gov. It later became arXiv and moved to Cornell University. Ginsparg has described this history in his own later writing, including a 2011 account. In 2013, Cold Spring Harbor Laboratory launched bioRxiv, a preprint server for the biological sciences, developed by John Inglis and Richard Sever. In 2019, medRxiv followed for the health sciences, as a collaboration involving Cold Spring Harbor Laboratory, Yale University, and BMJ. Other fields now have their own servers, but these three are the ones general readers are most likely to meet.
What you are usually looking at, then, is a complete draft that the authors chose to make public. Many servers assign a persistent identifier, keep a timestamp, and allow the authors to post later versions. That is part of the point: colleagues can read the work quickly, build on it, or criticize it, without waiting for the slower cycle of formal review and journal publication. Some preprints are later published in a journal, sometimes after substantial revision. Some remain preprints. Some are withdrawn. The document in front of you is one version, at one date. If the server shows v3, you are not necessarily reading the same text that first appeared as v1.
What the label does not do is equally important. A preprint has not automatically passed the kind of formal peer review that a journal organizes, in which editors invite outside readers to comment and an editorial decision is recorded. It has not been certified as reproducible. It has not been guaranteed free of errors in the analysis or the data. Readers may post comments, and specialists may notice problems quickly — that public scrutiny is one of the system's strengths — but informal comments are not the same as a structured review, and many preprints receive little public discussion at all. During periods of intense attention, such as the early COVID-19 pandemic, preprints made results available at high speed; they also made it easier for unverified claims to travel before any formal check had occurred. Speed and scrutiny do not rise automatically together.
Servers also differ in what they screen. A typical preprint platform may check that a submission is a scholarly manuscript and, in some fields, that it is not obvious nonsense or a dangerous medical prescription; that is not the same as judging whether the analysis is correct. Policies vary by server and have changed over time, so it is safer to read the platform's own description than to assume a uniform standard. The important constant is the one already stated: public posting is not, on its own, formal peer review.
For a careful reader, a few practical habits follow. Check the date and the version number, and look at whether the authors have posted a later revision or a withdrawal notice. Read the methods with the same attention you would give any other paper: a polished PDF is still only as strong as its design, sample, and analysis. If the work later appears in a journal, that later version may have changed, and it is the later text that should be quoted if you need the revised claims. None of this requires treating preprints as worthless. In several fields they are now a normal way to share work, and many important results first reached readers in that form. It only requires not confusing a public posting with a certificate.
A sentence that stays close to the facts would sound like this: the authors have released a dated manuscript that has not necessarily been formally reviewed yet. That is already useful information. It tells you to read the argument, not the badge, and to leave room for correction if a later version, or a later review, changes what the first posting seemed to show. In that sense a preprint is a claim placed on the public record, not a verdict on the claim.
More research explainers: https://www.crushsci.com
Top comments (0)