A preprint is a distinct publication state
Preprints make manuscripts public before or alongside journal publication. The screening performed by a server varies and is not the same as a journal's subject-matter peer review. Source evaluation should therefore separate the content, the platform's checks and the publication state. A preprint is not automatically unreliable, but deposit is not a blanket quality certificate either.
Assess authorship, affiliations, methods, data availability, competing interests and plausibility as you would for other research. Claims affecting health, safety or public policy need especially careful triangulation. Look for independent evidence and avoid presenting one early result as settled consensus.
Make the exact version recoverable
A server may host several revisions of the same manuscript. Data, analysis, author list and conclusions can change between them. Record the version number and date displayed for the file you used. Keep its DOI, landing page and access date with your notes rather than relying on a locally renamed PDF.
Read version notes and change statements. If no detailed record exists, compare the sections that support your argument. Do not quote version 1 while citing the metadata for version 3. When a later revision corrects an error, update the claim and any page, figure or section locator.
Connect the preprint to a journal version carefully
Search the title, creators and identifiers for a journal article. Some servers and metadata services display “published as” relationships. Verify the match; similar titles do not guarantee the same work. Compare the question, sample, principal results and conclusions rather than assuming the abstract is unchanged.
Where a journal version exists, it is often the natural citation for the final published account. The preprint still matters if version differences are being studied, it contains relevant material omitted later, or chronology is central. Cite both and explain the relationship. Do not imply that peer review independently validated every number; the published article still warrants critical reading.
Make status visible in prose and references
Include the document type in the reference list as required by the chosen style, commonly with a label such as “Preprint”. In the prose, a short phrase can establish the evidence boundary: “In a preprint that had not undergone journal peer review at the time of checking, the authors report …”. If a documented open-review process exists, describe it precisely rather than calling the work simply “peer reviewed”.
Pagination may change between versions. For HTML, use a stable section heading where the citation style permits it. Tie every quotation to the version actually read. The broader guide to identifying scholarly sources helps evaluate the publication and its methods beyond the status label.
Use an update log to prevent stale conclusions
Set a review date when first citing the work. Check the server record, DOI metadata, author profile and an appropriate subject database. If an article appears, compare more than the abstract. Revisit numbers, figures, exclusions and limitations on which your argument depends. Revise the surrounding interpretation when the evidence changes.
If a preprint is withdrawn or removed, record the date and public reason where available. Do not continue presenting it as unaffected evidence. When the withdrawal itself is relevant to your research, cite the status notice and preserved metadata transparently. Our retraction checking workflow provides the adjacent process.
A citation pattern without invented journal details
A generic structure is: “Author(s) (Year). Title [Preprint, version X]. Server. DOI.” Take each element from the version you inspected and apply the punctuation of the required style. Where the server has no explicit number, include the displayed posting date and access date. Do not manufacture volume, issue or page details simply to make the entry resemble a journal article.
Keep a local note of the status check, but cite the public record rather than your note. If your field requires a bibliography annotation, use it to state the version and review position succinctly. Consistency matters across the bibliography, footnotes and claims in the main text.