What Is a Preprint? Research Published Before Anyone Checks It

A preprint is the complete paper, written as the authors intend to submit it, posted on a public server with a permanent identifier and a date. It has not been reviewed by referees or accepted by a journal, so no external assessment of its methods or conclusions has occurred, which is the single fact that determines how it should be read. It is not a summary, a draft outline or an announcement, and it is not a lower standard of work, since the same paper typically appears in a journal later with revisions. Most journals now accept papers that have been posted as preprints, which was not true historically and was the main obstacle to the practice. The permanent record matters, since a posted preprint establishes priority and is citable, and the version history remains visible when the paper is revised.

The reasons are practical and most concern time and access:

• •Speed, since journal review takes months and sometimes years while posting takes a day
• •Priority, since the posting date establishes who had a result first
• •Access, since the paper is free to read for anyone rather than sitting behind a subscription
• •Feedback, since a wider audience can comment before the paper is finalised
• •Career timing, since hiring and funding decisions cannot wait for a journal schedule
• •Correcting the record, since a result that contradicts something widely believed can circulate immediately

The practice is far older than its recent prominence suggests. Physicists circulated paper copies of unpublished work by post for decades, and an electronic server established in 1991 formalised that habit, becoming the standard route for the field within a few years. Mathematics and computer science adopted it comparably early. Biology and medicine resisted far longer, on arguments about the risk of unreviewed medical claims reaching the public, and dedicated servers for those fields appeared only in the last decade. The pandemic accelerated the change enormously, since the need for immediate information outweighed the usual caution, and an enormous volume of work circulated as preprints and was reported on directly by journalists. That episode demonstrated both the value of speed and the specific hazards of unreviewed work reaching a general audience, which the field had predicted.

Widespread posting has changed the role of journals in ways that are still working through. The distribution function, which was historically the main thing a journal provided, has largely gone, since anyone can distribute a paper for free and permanently. What remains is certification, the stamp indicating that referees examined the work, and curation, the signal that a paper appearing in a particular venue met that venue's standards. Several initiatives have tried to separate those functions from publishing entirely, with services that organise peer review of preprints and attach the resulting reports to them, leaving the paper where it is. Funders in several countries now require that publicly funded research be freely available, which has accelerated the shift. The likely destination is contested, and the practical position for a reader is that the version of record and the version they can read may differ, and checking which one they have matters.

Treating a preprint appropriately requires the same scrutiny a reader should apply to any paper plus an awareness of what has not happened. Nobody has checked the methods, so obvious errors may be present that review would have caught, and the history includes cases where a preprint reported an implausible result that was withdrawn within days. The claims have not been moderated by an editor, so the language may be stronger than the evidence supports. Statistics have not been examined. Against that, peer review catches less than its reputation suggests, and reviewed papers contain errors regularly, so the difference is one of degree. The practical approach is to read the methods rather than the abstract, to check whether the conclusion depends on a small sample or a single analysis, and to treat a single preprint as one piece of evidence rather than as a finding, which is the correct approach to a single journal paper as well.

A complete paper posted publicly with a date and an identifier, before any referee has looked at it, which is what makes it fast, free and unchecked. Physics has worked this way since 1991 and biology and medicine only recently. The pandemic showed both the value of immediacy and the hazard of unreviewed claims reaching a general audience. Read the methods rather than the abstract, and treat one paper as one piece of evidence.