Preprint vs peer review, in short: a preprint is the author’s manuscript exactly as submitted, with no independent quality check, while the peer-reviewed published version has been read, challenged and often revised by named editors and anonymous expert reviewers before it appears in a journal. A 2026 scoping review synthesising 40 comparison studies found the two versions are usually similar in substance but differ in specific, measurable ways — disclosure statements, statistical framing, and the confidence of the conclusions.
A preprint is a complete draft of a research manuscript posted openly on a server such as bioRxiv or medRxiv before, or independently of, formal peer review at a journal.
Table of contents
- What is the difference between a preprint and a peer-reviewed paper?
- Is bioRxiv or medRxiv peer reviewed?
- What actually changes between preprint and publication?
- Is a preprint a reliable source — and is bioRxiv “a good journal”?
- Common questions about preprints and peer review
- What this means for research administrators and institutions
What is the difference between a preprint and a peer-reviewed paper?
A preprint has not been through editorial peer review. It has typically passed only a basic screening — checking that the file is a genuine research manuscript, scanning for plagiarism, and confirming author identity — not an assessment of scientific validity.
A peer-reviewed paper has been evaluated by a journal editor and, usually, two or more subject-matter reviewers who assess methodology, statistics, and the strength of the conclusions. The Bodleian Libraries at Oxford summarise this plainly: preprint servers verify scope and format, “but they do not check reliability or accuracy.”
The distinction is about process, not necessarily quality: a well-conducted study can be posted as an accurate preprint, and a flawed one can still pass review with revisions.
Is bioRxiv or medRxiv peer reviewed?
Neither bioRxiv nor medRxiv is peer reviewed. Both are preprint repositories, not journals. bioRxiv is operated by Cold Spring Harbor Laboratory for the life sciences; medRxiv is its clinical-research counterpart, run jointly with Yale University and BMJ. Both apply a basic screening process before posting, explicitly distinct from journal peer review, and both display a standard disclaimer that content has “not been certified by peer review.”
This is precisely why the ICMJE Recommendations ask authors to disclose, at the point of journal submission, whether a related preprint already exists — so editors and reviewers can see the unreviewed version alongside the manuscript under evaluation.
What actually changes between preprint and publication?
The most rigorous evidence comes from a 2026 scoping review in Research Integrity and Peer Review (Zoghbi et al.), which pooled 40 individual comparison studies published between 2019 and 2024, covering a median of 356 preprint-publication pairs per study (range: 19 to 73,256 pairs). Only 42% of preprints in the studies that tracked this were confirmed as eventually published in a journal (33 studies, interquartile range 22–67%).
| Element | Typical preprint state | Documented change after peer review |
|---|---|---|
| Title | Author’s original phrasing | Minor wording only — average lexical difference of ~0.2 |
| Abstract | Author’s original summary | 40–80% textual similarity retained across studies; lexical difference ~0.37, semantic difference ~0.16 |
| Conflict-of-interest statement | Often brief or absent | 17–23% of pairs gain a COI statement only in the published version |
| Funding disclosure | Sometimes omitted | 5–12% of pairs add funding disclosure only after review |
| Statistical methods | As originally analysed | Changed in 14–25% of studies, typically after reviewer request |
| Primary outcome reporting | As first framed by authors | Discordant between versions in 2.4–47% of study pairs, varying widely by field |
| “Spin” (overstated conclusions) | Present in 65.3% of preprints | Falls to 41.3% in the matched published version |
An earlier, widely cited paper in the same journal — Carneiro et al. (2020), cited more than 140 times — found that peer-reviewed articles had, on average, modestly higher quality of reporting than their preprint counterparts, though the absolute gap was small. Together, the two studies point to the same conclusion: peer review changes disclosure and framing more often than it changes the underlying data.
What rarely changes:
- The core dataset and raw results
- The overall study design and primary methodology
- The number of main figures and tables (though their content is sometimes revised)
Is a preprint a reliable source — and is bioRxiv “a good journal”?
bioRxiv is not a journal at all, so “is bioRxiv a good journal” is a category error — it is a repository with no editorial acceptance decision and no impact factor. The comparable question is whether a preprint hosted there is a reliable source, and the honest answer is conditional.
Given the scoping-review evidence above, most preprints are not wildly different from their eventual published form, but a meaningful minority carry disclosure gaps, unresolved statistical issues, or overstated conclusions that peer review later corrects. Readers citing a preprint should treat it as provisional: check whether a published version now exists, note that it has not been independently vetted, and flag its status explicitly when referencing it in grant applications, policy documents or news coverage.
Common questions about preprints and peer review
Has a preprint been peer reviewed?
No. A preprint has not undergone editorial peer review. Preprint servers check that a manuscript is genuine research and formatted correctly, but they do not evaluate its methodology, statistics, or conclusions — that scrutiny happens only after journal submission, during formal review.
What is the difference between preprint and peer review?
A preprint is the manuscript itself; peer review is the evaluation process a manuscript goes through at a journal. A preprint can exist with or without ever entering peer review, and the peer-reviewed published version is the same study after editors and reviewers have assessed and often revised it.
Is a preprint a reliable source?
Preprints can be reliable but carry more risk than peer-reviewed articles. Evidence shows most preprints change only modestly before publication, yet a documented minority show discordant statistics, missing disclosures, or overstated conclusions — so preprints should always be labelled as unreviewed when cited.
What are the disadvantages of posting a preprint before peer review?
The main disadvantages are reputational and interpretive: unreviewed errors can spread before correction, media outlets may misreport preliminary findings as settled, and negative public comments can occur before formal review even begins. These risks are why disclosure of preprint status at submission matters.
What this means for research administrators and institutions
Research offices increasingly need a documented position on preprints, not an informal one. The ICMJE Recommendations already require authors to declare related preprints at submission. UKRI permits preprints to be cited in grant applications and outputs reporting, provided they are clearly labelled as such. Crossref assigns DOIs to preprints under its “posted content” schema, giving institutions a citable, trackable identifier even before journal acceptance.
For research administration teams, this has practical consequences: tracking which preprints in a portfolio have since been published, checking whether reported authorship and contributor details stayed consistent between versions, and briefing researchers that a preprint’s conclusions are provisional until the peer-reviewed version confirms them. Institutions with active preprint policies should build these checks into standard research-administration workflows rather than treating preprints as an edge case.
The bottom line
Peer review does not typically rewrite a paper’s core findings. The documented evidence shows it most often tightens disclosure — conflicts of interest, funding, statistical methods — and moderates overstated conclusions, while leaving the underlying dataset and study design largely intact. Preprints and their published versions should therefore be read as two stages of the same evidentiary object, not two independent sources, with the published version carrying the added weight of independent scrutiny.
Leave a Reply