Why preprints are not peer reviewed yet

A preprint is a research manuscript posted on a public server before journal peer review has examined it, and its status is exactly that: not yet checked by reviewers. The biology server bioRxiv, launched in 2013, and its medical sibling medRxiv, launched in 2019, label every posting with notices that the work has not been certified by peer review. During the COVID-19 pandemic, preprints became a main channel for fast science, and studies of that period, including a 2022 analysis of COVID-19 preprint reproducibility, found the format's reliability varied widely.

This article is about how research gets published, not about any treatment; nothing here should guide a medical decision without a clinician's involvement.

What is a preprint, technically?

It is the authors' own manuscript, uploaded to a server that posts it within days after basic screening for plagiarism and non-scientific content. There is no selection for importance, no referee report, no required revision. Each version carries a permanent identifier and a date, and later versions sit beside earlier ones, so the record shows its own history. The manuscript may later be submitted to a journal, revised, and published after review, or it may never be reviewed at all.

The arrangement is deliberate. Preprints let researchers claim priority, share methods with colleagues, and invite comment while a submission winds through review, which commonly takes months. arXiv, the physics preprint server founded in 1991, normalized the practice for physical sciences decades before biology and medicine adopted it.

What does peer review actually add, later?

Reviewers read the manuscript against the data and the field's standards, raise objections, request analyses, and can recommend rejection. The process catches errors of method and interpretation often enough to matter, though studies of its reliability show it is a moderate filter, not a guarantee: it varies by reviewer, misses some flaws, and moves slowly. A journal publication therefore tells a reader that independent experts have at least examined the work. A preprint tells a reader that the authors chose to show it.

That difference in status, not in subject matter, is the point. Many preprints later appear in journals with modest changes; some are substantially revised after review; and some are withdrawn by authors or flagged by readers after errors surface. A 2022 study examining reproducibility of COVID-19 preprints found differences between preprint and published versions in a notable share of cases, which is the system working as designed, correction before certification.

Why did preprints spread so fast in 2020?

Because a pandemic could not wait for print. During 2020, researchers posted COVID-19 findings to medRxiv and bioRxiv at extraordinary volume, and journalists, clinicians and officials read them the same week. The speed was genuinely useful for sharing urgent evidence on transmission and treatment. It also placed unreviewed claims in front of audiences unaccustomed to checking status, and a handful of high-profile preprints were heavily revised or withdrawn after rapid public scrutiny, episodes later analyzed in the literature on preprint quality.

The episode left habits behind. Journals relaxed rules against citing preprints; funders and universities increasingly count them in evaluations; servers added prominent labeling. The format survived its stress test, with its trade-off, speed against certification, intact.

How should a reader treat a preprint finding?

As a claim with a timestamp. The work may be excellent; much of it is. But no independent expert has vetted it, so the ordinary reader-level checks matter more, not less.

1. Confirm the server label: posted before peer review is the default state.
2. Check the version and date, since later versions may revise or retract claims.
3. Look for a linked journal version, where review comments have already been addressed.
4. Treat the finding as preliminary in any retelling, and avoid changing behavior based on it alone.

How do we know when a preprint can be trusted?

The signals are indirect but readable. Trusted preprints tend to state funding and conflicts openly, share data and code, and describe methods in enough detail to be repeated. Rapid public commentary, posted responses and threads by researchers in the field, functions as informal review within days, though it lacks the structure of the journal process. Institutional endorsement helps less than readers assume: a famous affiliation certifies the authors' employer, not the manuscript.

The mature habit is proportional belief. A preprint consistent with prior work, from a team with relevant history, critiqued in public without collapse, deserves serious attention. A single preprint, unreviewed, contradicting a settled literature, deserves patience regardless of how often it has been shared.

Do preprints harm science?

They redistribute risk rather than create it. Unreviewed claims in public view can mislead, particularly when headlines strip the status label. But the same visibility lets the community find errors faster than any journal schedule allows, and the versioned record makes correction traceable. The harm comes less from the format than from reading preprints as finished findings, a habit the labels exist to prevent.

The label matters most where stakes are highest. In clinical medicine, a preprint suggesting a treatment benefit can reach patients before any reviewer has checked the statistics, so professional bodies have advised treating preprint evidence as provisional even in emergencies. In physics and mathematics, where preprint culture is decades old, readers apply the same caution by habit. The format itself is neutral; the obligation to state its status belongs to whoever passes the finding along.