A preprint is a complete draft of a research manuscript that a researcher makes publicly available on a preprint server before, or instead of, formal journal peer review. Preregistration is a related but separate practice: publicly time-stamping a study’s hypotheses, design, and analysis plan before data collection begins. Both are open-science practices aimed at making research more transparent and reproducible, and both now have their own DOI-issuing infrastructure — which is where confusion often starts, since a preprint DOI and a preregistration are not the same kind of record. This guide covers what each is, the main servers researchers use, whether either “counts” for funders and hiring committees, and how to cite and version them correctly.
What a preprint is — and isn’t
A preprint is a full manuscript — not an abstract, not a poster, not a working outline. It typically has not yet undergone the formal peer review a journal would apply, though a growing number of servers layer some form of screening or open review on top (see the comparison table below). Posting a preprint does not, by itself, constitute publication in a journal, and it does not substitute for peer review; it is best understood as a public, dated, citable version of a manuscript at a particular stage of its life. For the full operational definition and examples, see the Preprint dictionary entry and the broader explainer at Preprint Servers Explained.
Preprint servers by discipline
Preprint servers are not interchangeable. They differ by disciplinary scope, ownership, screening process, and what they do (or don’t) do with a manuscript once it’s posted. The table below summarizes the major servers; each has its own dedicated explainer linked from the table.
| Server | Primary scope | Operator | Screening before posting | Typical DOI |
|---|---|---|---|---|
| arXiv | Physics, mathematics, computer science, quantitative biology/finance, statistics, economics | Independent nonprofit (spun out from Cornell University, July 2026) | Author endorsement + moderator scope/format check (not peer review) | Yes, per version |
| bioRxiv | Life sciences | openRxiv (Cold Spring Harbor Laboratory) | Basic screening for plagiarism, non-scientific content, and biosecurity/ethics concerns | Yes |
| medRxiv | Clinical and health sciences | openRxiv, with Yale and BMJ partnership | Stricter screening, including a check for reporting patient-identifiable data and, for clinical trials, registration status | Yes |
| SSRN | Social sciences, economics, law, humanities | Elsevier | Editorial screening by subject-area networks | Yes |
| ChemRxiv | Chemistry | ACS, RSC, and other chemistry societies | Moderator review for scope and basic scientific soundness | Yes |
| PsyArXiv | Psychology | Center for Open Science (OSF-based) | Basic moderation via OSF Preprints infrastructure | Yes |
| Research Square | Multidisciplinary, plus in-review journal integration | Springer Nature-affiliated | Basic screening; offers optional preprint-linked peer review services | Yes |
Several of these overlap in scope (chemistry and biology work sometimes fits either bioRxiv or ChemRxiv, for example), and some journals now integrate directly with a preprint server as part of first submission. See ChemRxiv vs. bioRxiv for one worked comparison, and EarthArXiv for the earth-science equivalent not shown above.
Is arXiv peer reviewed?
No — and this is one of the most common misreadings of how arXiv works. Before a first-time submitter can post to a given arXiv category, they generally need endorsement from an existing arXiv author in good standing in that same subject area. Endorsement is explicitly not peer review: it checks that a submission is topically appropriate for the category, not whether its methods or conclusions are correct. Separately, arXiv moderators run an ongoing check for scope and format after submission, which can result in a paper being reclassified, held, or (rarely) rejected on scope grounds — again, not a scientific correctness review. See the arXiv Endorsement System entry and arXiv Preprints: What They Are and How to Use Them for the full mechanics.
Recent tightening (verified against arXiv’s own announcement): as of January 21, 2026, arXiv no longer accepts an institutional email address alone as sufficient grounds for endorsement of a new author. Per arXiv’s blog post announcing the change, a submitter must now either combine an institutional email with prior co-authorship on a paper already accepted to arXiv in the relevant endorsement domain, or obtain a direct personal endorsement from an established arXiv author in that domain. arXiv framed the change as a response to a rise in low-quality and automated submissions attempting to exploit institutional-email verification. This tightens who can post to arXiv; it does not change the underlying fact that arXiv postings, endorsed or not, are not peer-reviewed publications.
Do preprints count as publications?
There is no single answer — it depends who is asking and for what purpose, so treat this as three separate questions rather than one.
- For funders: Most major funders now explicitly recognize preprints as legitimate research outputs that can be cited in progress reports, grant applications, and funder-mandated output lists — NIH, for example, allows preprints to be cited in NIH grant applications and progress reports. This is not the same as treating a preprint as equivalent to a peer-reviewed article; it means funders acknowledge the preprint as a real, citable output of the funded work.
- For open-access compliance: Some funder OA policies now accept a preprint as a fallback compliant route when the peer-reviewed version can’t otherwise be made compliant — check the specific funder’s current policy language rather than assuming, since this detail varies and changes.
- For tenure, promotion, and hiring: This is the least standardized case. Some committees and disciplines (notably physics, and increasingly parts of biology and economics) treat a well-cited preprint as a legitimate scholarly contribution in its own right; many others, especially in fields where peer review remains the primary quality signal, do not count an unpublished preprint toward a formal publication record, or count it only with an explicit caveat. Where it matters for a specific evaluation, check the institution’s or committee’s own criteria rather than assuming either answer.
A related, practical question authors run into is whether posting a preprint jeopardizes later journal submission as “prior publication.” Most journals today explicitly permit prior preprint posting (a position formalized by many publishers and endorsed by COPE guidance for editors), but policies vary by journal and by field, and terms can attach (e.g., requiring a link back from the preprint to the published version). See Preprint Server Policies: Avoiding a “Prior Publication” Conflict and COPE’s Guidance on Preprints for Editors before submitting to a journal with an unclear preprint policy.
How to cite a preprint
A preprint citation needs the same core elements as a journal citation — author(s), title, year — plus two things a published-article citation doesn’t: the server name and, ideally, the version number, since a preprint can be revised after posting (see versioning below). Most citation styles now have an accepted format for this. The full style-by-style breakdown (APA, MLA, Chicago, and Vancouver) with worked examples lives at How to Cite a Preprint — use that guide for exact formatting; the short version is: cite the preprint server as the “publisher,” include the DOI if one exists, and note explicitly that the work is a preprint that has not completed peer review.
If the manuscript is later published in a peer-reviewed journal, cite the published version once it exists rather than the preprint, unless you specifically need to reference preprint-stage content (e.g., an earlier analysis that changed before publication) or the published version isn’t yet available.
Versioning and DOIs
Preprints are commonly revised after initial posting — in response to community feedback, further analysis, or in preparation for journal submission. Servers track this with explicit version numbers (v1, v2, v3…), each typically assigned its own DOI or DOI suffix, so a citation to “v1” and a citation to “v2” of the same preprint can point to meaningfully different content. Always cite the specific version you actually read or relied on, and check whether a newer version (or a published journal article superseding it) now exists before citing. See Preprint Version Numbers: Citation and Tracking Across Servers for the mechanics of how this works on different platforms.
Preregistration: what it is and how OSF’s workflow works
Preregistration means publicly and immutably time-stamping a study’s research questions, hypotheses, design, and (often) planned analysis before data collection begins, or before the data are examined. The point is to create a public, verifiable record that separates confirmatory analysis (testing a pre-specified hypothesis) from exploratory analysis (finding patterns after the fact) — a distinction that matters because presenting an exploratory, post-hoc finding as if it had been predicted in advance (sometimes called HARKing, “hypothesizing after results are known”) inflates the apparent strength of the evidence.
The Open Science Framework (OSF), run by the nonprofit Center for Open Science, is the most widely used preregistration infrastructure across disciplines. Through OSF Registries, a researcher completes a structured template (the specific fields depend on the registration template chosen — e.g., a general template versus a discipline-specific one), submits it, and the registration is then time-stamped and locked: it cannot be edited afterward, only supplemented with a dated addendum, which preserves the original record even if the study’s design later changes. The full step-by-step workflow is covered at OSF Preregistration: How to Register a Study on OSF Registries.
Preregistration is not a guarantee that a study was well designed or that its findings are correct — it doesn’t review the plan for quality. It only makes deviations from the original plan visible, which is what makes it a reproducibility tool rather than a quality-assurance one.
Registered Reports: preregistration plus peer review of the protocol
A Registered Report is a distinct, stronger publication format that builds on preregistration by adding formal peer review of the study plan itself, before the results exist. It runs in two stages:
- Stage 1: the author submits the introduction, hypotheses, and full proposed methods/analysis plan to a participating journal. This is peer reviewed on the merits of the question and design alone. If it passes, the journal grants in-principle acceptance (IPA) — a commitment to publish the eventual results essentially regardless of whether they turn out to be positive, negative, or null, provided the author follows the accepted protocol.
- Stage 2: the author carries out the study as registered, then submits the completed manuscript with results. Review at this stage focuses on whether the registered protocol was followed, not on whether the results were interesting or statistically significant.
This is the key difference from plain preregistration: preregistration is a self-filed, time-stamped record that no one reviews before you proceed; a Registered Report’s Stage 1 protocol is peer reviewed and conditionally accepted for publication before data collection starts. That combination directly addresses publication bias, since a null result registered at Stage 1 is just as publishable as a positive one. Registered Reports originated with a 2013 initiative at the journal Cortex and are now offered by several hundred journals across disciplines. For the mechanics of writing a Stage 1 protocol, see How to Write a Registered Report, and for the underlying format, Registered Reports: How Preregistered Study Plans Get Accepted Before Data Collection.
Zenodo: DOIs for datasets, software, and other outputs
Not every research output is a manuscript. Zenodo is a general-purpose open repository, operated by CERN and built on the OpenAIRE-supported InvenioRDM platform, that issues a DOI for essentially any research output: datasets, software releases, posters, presentation slides, negative results, protocols, and more. Two features make it a common default choice:
- No disciplinary or institutional gatekeeping — unlike a subject-specific preprint server or an institutional repository, Zenodo accepts deposits from any field and any researcher, which is part of why it’s frequently used for outputs that don’t fit elsewhere.
- Direct GitHub integration — a researcher can connect a GitHub repository to Zenodo, and each time a new GitHub release is published, Zenodo automatically archives a snapshot of that release and mints it a citable DOI. This is the standard way to make a specific, citable, permanent version of a software release (as distinct from the repository itself, which keeps changing) — useful for citing “the exact code used to produce Figure 3” rather than “whatever is currently in the repository.”
A Zenodo DOI is not a peer-review credential and doesn’t imply the deposited content was checked for correctness; it’s a persistent-identifier and archiving service. For how Zenodo compares to other repository options for a given deposit, see Zenodo vs. OSF, Zenodo vs. Dryad, and Zenodo vs. Figshare.
Frequently asked questions
What is a preprint, in one sentence?
A preprint is a complete research manuscript posted publicly on a dedicated server before, or independent of, formal journal peer review.
Does posting a preprint stop me from publishing in a journal later?
Usually no — most journals now explicitly permit prior preprint posting, but always check the specific target journal’s policy, since terms vary. See Preprint Server Policies: Avoiding a “Prior Publication” Conflict.
Is a preregistration the same as a Registered Report?
No. A preregistration is a self-filed, time-stamped record of your plan that no one reviews before you proceed. A Registered Report’s Stage 1 protocol is peer reviewed by a journal and conditionally accepted for publication (in-principle acceptance) before data collection begins.
Do I need a Zenodo account to get a DOI for my dataset or code?
Yes, a free Zenodo account (via GitHub or ORCID login) is required to deposit content or connect a GitHub repository for automatic release archiving.
Can I cite a preprint version that was later revised?
Yes, but cite the exact version you used (v1, v2, etc.) since content can change between versions, and check whether the work has since been published in a peer-reviewed journal. See Preprint Version Numbers.
Last verified: August 16, 2026. The arXiv endorsement policy change is sourced directly from arXiv’s own January 21, 2026 blog announcement.
