Preprints, Accepted Manuscripts and Versions of Record: Which One to Cite

A proceedings paper may exist as a preprint, an accepted manuscript and a final published version. Cite the version you actually read, and prefer the published one when it is available.

Last updated 6 November 2026

Version terms from NISO

NISO, a standards body, and ALPSP, a publisher association, issued the recommended practice RP-8-2008 on journal article versions. It defines seven stages, and the terms are now widely used for conference papers too. NISO has since started work to update the practice.

• Author's Original (AO): the version the author prepares before submission.
• Submitted Manuscript Under Review (SMUR): the version under peer review.
• Accepted Manuscript (AM): the version accepted for publication, before the publisher's formatting.
• Proof (P): the typeset version sent for checking.
• Version of Record (VoR): the final published version.
• Corrected Version of Record (CVoR): the VoR after a correction.
• Enhanced Version of Record (EVoR): the VoR with added features, such as linked data.

A preprint is not one of the NISO terms. In practice it usually means a version shared before or outside formal review, and it often corresponds to the AO or SMUR stage.

Which version should you cite?

The general principle is to cite what you read and to be able to prove it. Beyond that, these rules of thumb apply.

1. If the final published version exists and you can read it, cite it. Page numbers, volume details and the DOI belong to that version.
2. If you can read only a preprint or accepted manuscript, cite that version and say so, using the format your style guide gives for preprints.
3. If you rely on a specific claim, check it in the final version, since the text may have changed in review.
4. If your venue or supervisor sets a rule, follow it.

Check the author instructions of the venue where you publish. Some ask for the published version, while others accept a stable preprint. The APA guide and the IEEE guide on this site show how each style records the proceedings paper.

A simple habit helps here. When you save a PDF, record where you got it, the date, and the version label on its first page. Many preprints and accepted manuscripts carry a note such as accepted for publication or a version number. This record lets you cite accurately later, and it lets you switch to the final version if one appears before you submit your own work.

arXiv and numbered versions

arXiv labels each submission as v1, v2 and so on. According to its help pages, any change to the paper creates a new version, and earlier versions stay publicly available. The exceptions are small additions to the record, such as a journal reference, DOI or report number.

When you cite an arXiv paper, include the identifier and, where it matters, the version number. If the work was later published in proceedings, check the arXiv record, because authors often add the published reference there. The two texts may differ, so do not assume they match. See identifiers for proceedings for how these labels fit with DOIs.

Zenodo and versioned DOIs

Repositories such as Zenodo support versions of the same deposit. Zenodo's help pages state that a new version is a separate record with its own metadata, files and persistent identifier, linked to the other versions. A reader who cites a specific version can therefore rely on its files staying the same.

Zenodo also provides a DOI that stands for the whole set of versions, often called the concept DOI. Zenodo's current help pages describe how this works, so confirm the exact wording there. The general rule holds: cite the version DOI for exactly what you used, and the overall DOI only when you mean the work in general.

Corrections, updates and retractions

Published papers can change after publication. A publisher may issue a correction, an addendum or, in serious cases, a retraction. Crossref runs a service called Crossmark, which gives readers a button that shows whether an item has been updated, corrected or retracted. Participating publishers commit to registering updates and to keeping an update policy.

Not every publisher takes part, and many conference series do not. Absence of a Crossmark button does not prove that a paper is unchanged. Before you cite an important paper, search the title with the words correction or retraction, and check the publisher's page and the DOI landing page.

If you cite a paper that was retracted, say so in your text. Do not cite it as sound evidence. The verification guide covers metadata checks.

Keep your own records too. If you cite a paper in a thesis that will take months to finish, repeat the check just before submission, since a correction may appear in the meantime. Note the date of your last check in your notes.

Common version mistakes

• Citing the preprint's page numbers for the published paper, or the reverse.
• Giving the published DOI for a text that you read only as a preprint.
• Citing the event year when the published year differs.
• Citing a link to a live preprint without noting the version or date.
• Missing a later correction that changes a result.

The list of common mistakes gives more detail on related errors. If you use a reference manager, check the item type and the version, since automatic imports often pick the preprint record.

Frequently asked questions

Sources and further reading

Related topics

• DOI, ISSN, ISBN, URN and ORCID: Identifiers for ProceedingsWhat a DOI, ISSN, ISBN, URN and ORCID iD each identify, how to check them, how Crossref and DataCite differ, and what to do when a paper has no DOI.
• How to Verify That a Citation Is Real and CorrectStep-by-step check that a reference is real and correct: search the title, resolve the DOI, compare Crossref and DBLP, and spot fabricated or AI citations.
• Ten Common Mistakes When Citing ProceedingsTen practical errors in citing workshop and conference papers, from a missing proceedings title to the wrong year or version, with a fictional example.