Preprint vs Accepted Manuscript vs Version of Record

Your paper exists in several versions, and mixing up the preprint, accepted manuscript and version of record causes citation and policy problems.

Explainer · 6 min read · Directive Publications Editorial Team

The three versions (plain language)

1. Preprint — your manuscript before peer-reviewed acceptance (often posted on a server)
2. Accepted manuscript (AAM) — peer-reviewed, accepted, not yet final publisher formatting
3. Version of Record (VoR) — the final published article of record (DOI, typesetting, permanent home)

The preprint is whatever you posted publicly before a journal accepted the work. It is timestamped, it usually gets its own DOI, and it can have versions of its own — v1, v2, v3 — as you update it. What it does not have is a peer review decision behind it. Claims, numbers and even whole analyses can change between the preprint and acceptance, which is exactly what review is for.

The accepted manuscript is the precise file the editor accepted, including your final revisions and nothing after that. It normally looks like your own document: figures sometimes still at the end, references in whatever style you submitted, no journal branding, no publisher page numbers. Repositories usually want this version, because it carries the peer-reviewed content without the publisher's typesetting.

The version of record is the article as the journal publishes it — copyedited, typeset, given final pagination or an article number, and carrying a DOI that resolves to the publisher's page. It is also the version that later notices attach to. Any correction, expression of concern or retraction is fixed to the VoR, not to the copies floating around elsewhere.

If someone sends you a PDF and you are not sure what it is, three signals settle it quickly: journal branding with continuous page numbers means VoR; a bare manuscript with figures at the end and no publisher logo means AAM; a server cover sheet with a version number and posting date means preprint.

Why it matters

• Citing: Prefer the VoR when it exists
• Sharing: Check license and journal policy for each version
• Correcting: Corrections usually attach to the VoR

Citing. The rule is short. If a DOI resolves to a published article, cite that. Cite a preprint only when no published version exists yet, and when you do, make the version visible in the reference itself: the word "preprint", the server, the posting date, and the version number you actually read. What goes wrong is not abstract. You quote a sentence that reviewers made the authors delete, or a coefficient that moved during revision, and your reader follows the citation to the VoR and cannot find what you attributed to it. In a review or meta-analysis the failure is larger: counting a preprint and its published version as two sources double-counts the same data.

Sharing. Ask three questions before you post anything: which version, where, and when. The licence you chose when the preprint went up governs the preprint. Your publishing agreement governs the accepted manuscript, which is often allowed in an institutional or subject repository, sometimes only after an embargo period you need to look up. The VoR is normally redistributable only if it was published under an open licence. The publisher PDF is the tidiest file and the one people instinctively reach for, and it is usually the one you are not permitted to upload to a scholarly networking site. Emailing the accepted manuscript to a reader who asks is generally the safer response.

Correcting. The correction attaches to the VoR, and no other copy inherits it. That part is your job to fix by hand. Go back to the preprint record and add a note pointing to the VoR and to the correction notice, and ask your repository to do the same on its copy. Do not quietly swap a file for a fixed one: servers keep every version anyway, so a silent replacement is both detectable and worse-looking than a labelled update.

Start from Academy guidance via Directive Academy (article versions explained) and open-access policy pages on Directive Publications.

Practical habits

• Label files clearly (preprint, accepted, vor)
• Update the preprint with a link to the VoR after publication
• Do not treat a preprint finding as “published peer-reviewed” in promotion files without clarity

Name files so the version is readable without opening them. One folder per manuscript, and a name that carries author, year, short title and version: nakamura-2026-soil-carbon_preprint-v2.pdf, nakamura-2026-soil-carbon_accepted-2026-03-11.pdf, nakamura-2026-soil-carbon_vor.pdf. This prevents a very ordinary failure: a co-author asks for "the paper", you attach the newest thing in your downloads folder, and it turns out to be a tracked-changes revision or a superseded preprint.

Save the accepted manuscript on the day it is accepted. Assemble the exact accepted text and figures, name the file, and put it in that folder immediately. Months later, when a funder or repository asks for the AAM, rebuilding it from a chain of email attachments is slow, and you can easily send something that is not quite what was accepted.

Close the loop when the VoR appears. Add the link to the preprint record, then to your ORCID entry, your group's website, your CV, and any repository deposit. Every stale copy without a forward link is a reader who stops at an outdated version.

Be exact in promotion and assessment files. Keep preprints under their own heading, labelled as preprints, and move an entry into the peer-reviewed section only once it is accepted. Never list both the preprint and the published article as separate outputs. A panel member who spots the same work counted twice will read the rest of your file far more sceptically.

Situations that catch people out

You find an error after acceptance but before publication. Tell the editor at once, while the article is still in proof. A change made at proof stage is an ordinary production edit; the same change made after publication needs a correction notice attached to the record permanently.

Your preprint is being cited and the paper has since changed. You cannot withdraw citations, but you can make the current version easy to reach. Post an update to the preprint record saying the peer-reviewed version supersedes it, give the DOI, and state plainly if a headline number changed.

A collaborator sends you the publisher PDF and asks you to post it. Do not assume their permission covers your posting. Check what the destination allows for that version, and default to the accepted manuscript in a repository with a link to the VoR.

A reader or journalist quotes your preprint as a published finding. Correct it early and specifically: name the version they read, say whether the finding survived review unchanged, and give the DOI of the published article if one exists.

Bottom line

Versions are not interchangeable. Know which one you are posting, citing, and correcting — then point readers to the VoR when available.

Most version problems are cheap to prevent and expensive to unwind. Ten seconds of file naming, one link added to a preprint record, and one honest label in a CV save you a correction request, a takedown notice, or an awkward conversation with an assessment panel.

Open Access at Directive · Submit