A single piece of research now commonly exists in three or four forms at the same time: a preprint posted before review, an accepted manuscript that has passed review but not yet been typeset, and the final published version of record — sometimes with a later corrected or updated version on top. Readers, and even authors, routinely confuse them, and citing the wrong one can misrepresent what was actually validated. This article sets out what each version is and how peer review sits between them. It builds on the broader taxonomy in the research-outputs domain and pairs with the side-by-side explainer at preprint versus published article.
The versions, in order
The preprint
A preprint is a complete research manuscript posted to a public server — arXiv, bioRxiv, medRxiv, SSRN, and many others — before, or in parallel with, formal peer review. Its defining feature is speed and openness: it makes findings available immediately and citable via a persistent identifier, usually a DOI, without waiting for a journal’s review cycle. Its defining limitation is the flip side of the same coin: a preprint has not been through independent peer review, so its claims have not been externally vetted. A preprint is a legitimate, citable output — not a lesser draft — but it carries a different epistemic status from a reviewed article, and that status must be made clear wherever it is used.
Peer review: the step between
Between the preprint and the published article sits peer review — independent evaluation by qualified reviewers, organised by a journal editor, which may accept, reject, or (most often) require revision. Peer review is not a guarantee of correctness; it is a quality-control and improvement process. What it changes is the manuscript’s standing: a reviewed and accepted article carries the journal’s editorial endorsement that the work met its standards, which a preprint does not. Understanding this is the key to the whole picture — the versions differ mainly in what has happened to them, and peer review is the event that separates the unreviewed preprint from the validated article.
The accepted manuscript (postprint)
Once peer review concludes and the journal accepts the paper, the author’s final reviewed-and-revised file is the accepted manuscript, often called the postprint or author-accepted manuscript (AAM). It contains the intellectual content that passed review but lacks the publisher’s copy-editing, typesetting, and final pagination. The accepted manuscript is the version most commonly self-archived in institutional repositories under green open access, frequently after an embargo. It is content-equivalent to the published article in its claims, but it is not the citable, formatted final object.
The version of record
The version of record (VoR) is the final, published, formally citable version: copy-edited, typeset, paginated, assigned its DOI, and lodged with the publisher as the authoritative instance of the work. It is the version the scholarly record points to, the one that carries any later corrections or retractions, and the one that should normally be cited. The concept of a version of record exists precisely so that, among several coexisting forms, there is one designated authoritative object that the record and its corrections attach to.
How they fit together
The clean way to hold this in mind is as a sequence of states of one work:
- Preprint — complete, public, citable, not peer-reviewed.
- (Peer review happens.)
- Accepted manuscript / postprint — peer-reviewed content, not yet publisher-formatted; the usual green-OA archive copy.
- Version of record — the final, formatted, authoritative, citable version.
Crucially, these can all exist simultaneously and should link to one another. A well-managed preprint server displays a link from the preprint to the published version of record once it appears; the version of record, in turn, may acknowledge the preprint. Persistent identifiers are what make this linkage reliable: the preprint and the VoR each have their own DOI, and the relationship between them is recorded in metadata so that a reader arriving at one can find the other.
Which version to cite
- Cite the version of record where it exists. It is the authoritative, corrected, formally published instance, and citing it ensures your reference points to what was validated and to any subsequent corrections.
- Cite the preprint as a preprint when that is genuinely what you used — for example a result not yet published elsewhere — and label it clearly as a preprint, with its DOI, so a reader knows it has not been peer-reviewed.
- Do not cite a preprint as though it were the published article. If a version of record now exists, prefer it; the preprint and the final version can differ in their conclusions after revision.
- Check for a newer version. Preprints are often updated; the VoR may carry corrections. Cite the specific version you relied on, and prefer the most authoritative current one.
A note on what preprints do and do not change
Preprints have made research faster and more open, and they are now a first-class part of the scholarly record rather than a fringe practice. But they do not replace peer review or the version of record; they sit before them. The healthiest reading of the current landscape is not preprint versus journal but a pipeline in which the same work moves from open-but-unreviewed to reviewed-and-authoritative, with each stage clearly labelled and linked. Confusion arises only when the labels are dropped — when a preprint is presented, or cited, as if it had the standing of the version of record.
Where shared vocabulary fits
“Preprint”, “postprint”, “accepted manuscript”, and “version of record” are used inconsistently — and sometimes interchangeably — across servers, repositories, and citation styles, which is exactly how the wrong version ends up cited. A shared, federated vocabulary that defines these versions precisely and records the relationships between them is what lets a citation point unambiguously to the right object. Supplying that definitional layer is the role the CASRAI dictionary is designed to play; the relevant terms sit in the research-outputs domain.