Preprints vs. Peer-Reviewed Papers: Does It Matter Which One You Cite?
A decade ago, citing a preprint in an academic paper could raise eyebrows. Today it's routine, especially in fast-moving fields like machine learning, physics, and parts of biomedicine, where the gap between "results exist" and "results are formally published" can stretch past a year. But routine isn't the same as risk-free, and knowing when a preprint is a reasonable citation versus a liability is worth being deliberate about.
What "preprint" actually means
A preprint is a manuscript posted publicly before, or sometimes instead of, formal peer review. Platforms like arXiv, bioRxiv, medRxiv, and SSRN host them by the millions, and many eventually get published in a journal after revision. Some never do, either because the authors never submitted them for review or because the paper didn't survive it. The preprint you're looking at might be six months away from a polished journal version, or it might be the final form the work ever takes.
Why fields treat them so differently
In computer science and physics, preprints are often the primary way results circulate; by the time a paper appears in a journal or conference proceedings, the community has usually already read, cited, and built on the arXiv version for a year or more. In clinical medicine, the norm is much more conservative, since an unreviewed claim about a treatment or diagnostic tool carries real downstream risk if it's wrong. A citation practice that's completely normal in one discipline can look sloppy in another, so the safest approach is to follow the convention of the field you're writing in rather than assume a blanket rule.
Questions worth asking before you cite one
Has it been sitting unpublished for an unusually long time? A preprint that's been up for three years with no journal version is a mild warning sign — it may have been rejected, revised heavily, or simply abandoned.
Does the claim matter to your argument? A preprint used for background context or motivation is lower stakes than one your central conclusion depends on. Save extra scrutiny for the load-bearing citations.
Can you find the peer-reviewed version instead? Many preprints have a published counterpart with a different title or an added co-author. A quick DOI or title search before you finalize your reference list can upgrade a preprint citation to a peer-reviewed one for free.
Is it from a preprint server with any screening at all? Servers like bioRxiv do a light sanity check (no plagiarism, not obviously unscientific) but this is nowhere near peer review, and it's worth remembering that distinction rather than reading "hosted on a reputable server" as "vetted."
None of this means avoid preprints. It means read them as what they are: research the authors are confident enough to put their name on, not research a field has yet agreed is correct. When Research Verifier surfaces a result from arXiv alongside one from a peer-reviewed database, it shows citation count and publication year for both, so you can weigh a preprint against the rest of the evidence yourself instead of treating every result as equally settled.