Preprints, versions, and which one gets cited

Does posting a preprint count as prior publication?

Not on its own. The ICMJE does not treat a preprint posted on a qualifying server as prior publication. It does note that "posting of work as a preprint may influence a journal's interest in or priority for peer review."¹

Individual journals set their own policy within that, and some are stricter than others. Reading the journal's instructions before posting rather than after is the version of this problem that has an easy answer.

The distinction that matters is between a preprint and a duplicate submission. The ICMJE is direct about the second: authors "should not submit the same manuscript, in the same or different languages, simultaneously to more than one journal."¹ Its stated reason is that duplicate submission causes editorial disputes and duplicates the work of peer review.¹

Do I have to tell the journal about the preprint?

Yes, and with a link. The ICMJE states that authors should inform a journal where the submitted work has been posted on a preprint server, and should provide a link to it.¹

The obligation continues after acceptance. It is the authors' responsibility to ensure that preprints are amended to point readers to subsequent versions, which usually means adding a note and a DOI to the preprint record.¹

Authors forget that second step regularly, which is how a superseded preprint goes on collecting citations for years after the peer-reviewed version appeared. The correction takes a few minutes on the server's own interface.

Disclosure at submission is cheap and the alternative is not. A journal that discovers an undisclosed preprint mid-review is dealing with a question about candor rather than about policy.

Can a preprint be taken down?

Generally not. bioRxiv, which is owned and operated by Cold Spring Harbor Laboratory, states that papers cannot be removed once posted, because each receives a DOI and becomes part of the scientific record.²

What exists instead is a withdrawal statement. Authors who no longer stand behind their findings, or who have found a fundamental error, can post one, and the preprint remains visible with that notice attached.²

Removal has happened in rare cases for legal or safety reasons.² It is not something to plan around. A preprint is a publication in the sense that matters here, which is that it cannot be unpublished.

That permanence is the argument for reading a preprint once more before posting it. The version that goes up is the one that will be indexed, cited, and found by people who never see the journal article.

What screening does a preprint go through?

Screening rather than review. bioRxiv states that manuscripts are "not certified by scientific peer review, edited, or typeset before being posted online," and that they may therefore contain errors or omissions.²

There is a check before posting. Submissions pass a multi-step screen covering plagiarism, non-scientific content, and material that could endanger public health, including dual-use research.²

The absence of copyediting is the part relevant to an author's decision. Whatever is posted is what readers see, in the state it was uploaded, for as long as that version exists.

Screening also has a schedule. Posting is fast compared with review, but it is not instant, and a preprint timed to appear alongside a conference talk needs a few days of margin.

Should a preprint be edited before it goes up?

That is a judgment about audience rather than about rules, and the case for it is that the preprint is often the version most people read. It is free, it is indexed, and it appears months before the journal version.

Authors weighing the cost usually decide by how far the paper is from submission. A manuscript going to a journal next week gains little from a separate pass, since the revision will change it. A preprint that will sit unchanged for six months while a review runs is a different case.

Language faults in a preprint also travel. Reviewers frequently read the preprint, and a sentence that confused a reader in March is available to confuse a reviewer in June.

Which version do people cite?

The most recent one, by convention. The ICMJE states that the citation should be to the most recent version. It also asks that earlier electronic versions prominently note that more recent versions of the article exist.³

Preprint servers implement that with version numbering rather than replacement. bioRxiv posts revisions under the same DOI, with earlier versions still reachable through the record's history.² Authors can revise at any point up to journal acceptance.²

In practice authors have less control over this than they would like. A reader who found version one has no reason to check for version three unless the record tells them, which is the whole point of the notice requirement.

Citation managers make the problem worse rather than better. A reference imported from a preprint in 2024 stays a preprint reference in 2026 unless someone refreshes it.

What about copyright and licenses?

Preprint servers ask the author to choose. bioRxiv authors retain copyright and select from six license options, ranging from one requiring permission for any reuse to a public domain dedication.²

The choice interacts with what a journal will later accept, and journals differ on which licenses they can work with. Checking that combination before posting avoids a conversation at the acceptance stage.

An editor does not choose a license for you, and this is not a question we answer. Where a manuscript carries a license statement, we can check that the statement is consistent between the preprint, the submission, and the funder's requirements as you have described them.

Funder mandates add a third set of conditions on top of the server's and the journal's. Those come from your institution's research office rather than from an editor.

What can an editor do across several versions of one paper?

Consistency across them is the job, and it is harder than it sounds once three versions exist. Author lists, affiliations, funding statements, data availability statements, and competing interest declarations all drift between a preprint and a final submission. None of that drift is deliberate, and none of it is visible without a comparison.

A comparison pass across versions catches the coauthor added in July who is missing from the preprint, the grant number that changed, and the abstract that no longer matches the results.

Reference lists are the other place versions separate. A preprint citing a colleague's own preprint often needs updating once that colleague's paper appears, and nobody notices unless someone checks.

Every document a client sends us is treated in confidence. That includes an unposted preprint, which is an unpublished manuscript until the moment it goes up, and is handled as one.