Short answer to what is a preprint medical research: a complete public draft of a study, not yet certified by peer review. Treat its findings as provisional, because results and conclusions may change before any journal version appears.
The NIH describes preprints as complete and public drafts of scientific documents. They now surface regularly in medical news today, often without the context readers need. This guide explains how a preprint is posted, what screening does, how PubMed labels preprints, and how to judge a preprint claim in a news story.
How a preprint is posted and what it is not
A preprint is a full manuscript made public before peer review. The JAMA Network definition, as quoted by Cabells, calls it a full and complete version of a research manuscript made publicly available. It is not an abstract, a press release or a conference poster. The authors have finished a draft and chosen to share it.
The NIH says that certification by peer review is the key distinction between a preprint and an accepted author manuscript or a published article. For how that certification works and where it falls short, see what is peer review in medical journals.
A typical path runs in stages:
- Drafting. The authors prepare a complete manuscript, with methods, results and conclusions.
- Posting. They submit it to a preprint server such as medRxiv or bioRxiv. These are websites that host preprints.
- Screening. The server checks the submission against its public screening policies, including scope, before the draft goes live.
- Marking. The public record is labelled as not certified by peer review.
- Revision. Authors may revise preprints and post updated versions to the server.
- Journal publication, if it happens. A journal may later publish a version after its own review, and the preprint record can link to it.
Because the document looks like a journal paper, the missing certification is easy to overlook. The labelling and screening rules below exist to prevent that.
medRxiv, bioRxiv and screening checks
medRxiv and bioRxiv are preprint servers. Both are named by the NIH among the servers (alongside arXiv and Research Square) eligible for Phase 2 of its Preprint Pilot. Eligibility comes with obligations that tell readers what a server’s label means.
| Server requirement | What it means | Planning use |
|---|---|---|
| Screening process posted publicly, with scope policies | The server explains what it checks and what it accepts | Read the server’s own screening page before assuming a posted draft has been vetted for anything beyond scope |
| Records clearly marked as not certified by peer review | The not-peer-reviewed notice is a standing requirement, not a courtesy | If a news story or social post omits the notice, check the record itself |
| Links from the preprint to the published journal article, when available | The server record is the first place a later publication should be signposted | Look for the link on the preprint page before searching elsewhere |
These requirements come from the NIH Preprint Pilot documentation.
Screening is a gate on what gets posted, not a verdict on whether the findings are right. It checks things such as scope and policy fit. Peer review is a separate certification, and a preprint that passes screening still carries the not-certified label.
Screening can be strict where harm is plausible. Cabells reports that bioRxiv screening now blocks most research that makes treatment recommendations for COVID-19. That shows servers can limit risky claims, but a reader should not assume every server applies the same filters.
How PubMed handles preprints
PubMed carries preprints in a way that keeps the warning attached. The NIH says PMC and PubMed show a prominent panel noting that the article has not yet been peer reviewed. A preprint indicator has also been added to the citation metadata and the Cite tool, and it appears in PMC and PubMed search results.
Every preprint record in PMC and PubMed links to the complete record on the preprint server’s own website. A PubMed listing is therefore a pointer, not an endorsement. Use the label in the search results to tell a preprint from a journal article at a glance.
Not every preprint is in PubMed, because the NIH inclusion is limited by scope. Phase 1 of the pilot made more than 3,300 preprints reporting NIH-supported COVID-19 research discoverable between June 2020 and June 2022. Phase 2 launched on January 30, 2023. It covers preprints that acknowledge direct NIH support or have an NIH-affiliated author, posted from January 1, 2023. Once a week, the National Library of Medicine identifies in-scope preprints. A preprint missing from PubMed may simply fall outside that scope.
Should you trust preprint results?
Treat preprint results as provisional leads, not settled findings. The NIH states that any aspect of the research, including the results and conclusions, may change as a result of peer review. A headline number can survive review, shift or disappear.
Preprints are not junk, and they are not reliable by default. Cabells notes that over 30,000 of more than 125,000 COVID-19 manuscripts began as preprints, so early evidence often arrives this way. The same source describes a COVID-19 preprint that was widely read before it received hundreds of critical comments and was eventually withdrawn. It also warns that laypeople may place unjustified trust in unreviewed results.
The practical judgement depends on the claim:
- Low stakes, exploratory finding. A preprint is fine as a signal of what researchers are working on, if the story says plainly that it is unreviewed.
- Advice for patients or treatment choices. Wait for the journal version or other evidence before changing anything. A preprint should not drive a health decision on its own.
- Drug approval claims. A preprint cannot settle these. Regulatory decisions follow a separate process, set out in our guide to how does the FDA approve a new drug.
Checking for a later published version
Check the preprint server record first, then PubMed, then the journal. This order follows how the links are meant to work.
- Open the preprint page. Eligible servers should provide a link from the preprint to the published journal article, when available. A link there is the clearest sign that a journal has published a version.
- Look for revised versions. Authors may post updated versions, so confirm which version the story describes. A story may quote an older draft.
- Search PubMed. Look for the title or authors and read the label. A preprint indicator means the record is still the unreviewed draft. A journal citation means a published article exists.
- Compare the main claim. If a journal version exists, check whether the headline result, the conclusion and the caveats match what the news story reported.
If no published version turns up, that is not proof the work is flawed. It may still be under review, or the journal version may not yet be linked. The right conclusion is to keep treating the claim as unreviewed.
A rulebook for news stories that cite preprints
A fair story names the preprint, says it has not been peer reviewed, and signals that findings may change. Readers can hold stories to these expectations:
- Label the source. The word “preprint” or an equivalent plain description should appear near the finding, not only in a footnote. The NIH panel uses the wording that the article has not yet been peer reviewed.
- Avoid certainty verbs. Phrases such as “proves” or “shows conclusively” do not fit an unreviewed draft. Hedged wording such as “suggests” or “reports” does, and the same applies to “early findings”.
- Say what could change. Because results and conclusions may change after review, a good story says the numbers are preliminary.
- Link the record. The preprint server page lets readers see the notice, the version and any link to a published article.
- Update when a journal version appears. A story that ignores a later publication, or a withdrawal, leaves readers with an outdated claim.
Quality varies across the field, so a preprint label alone says little about whether a result is strong. It does say how much weight to give the claim today: provisional, pending review and open to revision.
Frequently asked questions
What is medRxiv?
medRxiv is a preprint server, a website where authors post manuscripts before peer review. The NIH lists it among the servers eligible for Phase 2 of its Preprint Pilot, together with bioRxiv, arXiv and Research Square.
Does a PubMed listing mean a study was peer reviewed?
No. PubMed shows a panel stating that a preprint has not yet been peer reviewed and adds a preprint indicator to search results and citation data. The listing links back to the preprint server’s record.
Can a preprint change after it is posted?
Yes. Authors may revise a preprint and post updated versions to the server, and peer review may change the results and conclusions in any later journal version.
Where does a preprint record point to a published article?
Eligible servers should provide a link from the preprint to the journal article when one is available, so the server record is the first place to look.
