Try Wonders for free
Preprints, working papers and other non-refereed work are part of the scholarly record, especially in fast-moving fields. Citing them correctly comes down to one thing: making the reader able to see what status the work has, and which version of it you read.
Why These Sources Need Special Treatment
When you cite a peer-reviewed article, a reader can assume it passed editorial and referee scrutiny. A preprint has not been through that. That does not make it wrong — but it makes the label part of the citation, not an optional extra.
Types of Non-Peer-Reviewed Sources
Preprints are complete manuscripts posted publicly before, or instead of, peer review. The main servers are arXiv (physics, maths, computer science), bioRxiv and medRxiv (biology and medicine), SSRN (social sciences and law), PsyArXiv (psychology) and OSF Preprints.
Working papers are in-progress research circulated through an institutional series — the National Bureau of Economic Research, the World Bank, university departments. Some become journal articles; some do not.
Conference papers vary from fully refereed to accepted on an abstract. Find out which, because it changes what the citation means.
Theses and dissertations are examined by a committee rather than refereed by a journal. They are generally treated as credible, with their own source type.
Technical reports from agencies, companies and research institutes also sit outside journal peer review.
General Principles Across All Styles
Whatever style you use:
- Identify the source type. Do not let a reader assume peer review that did not happen.
- Name the repository or series.
- Give the version or revision, where the server provides one.
- Give the posting or last-revised date.
- Use the DOI where one exists.
And: if the preprint has since been published, cite the published version — unless you are specifically discussing the preprint, or the differences between the versions.
Quick Reference by Major Style
These are templates. Fill in your own source's details; the exact element order for your edition belongs to your style's reference-examples page.
APA treats a preprint as its own source type, with the repository standing where a journal name would:
[Surname], [Initials]. ([Year]). [Title of preprint] [Preprint]. [Repository name]. [DOI]
MLA puts the platform in the container position:
[Surname], [First name]. "[Title of Preprint]." [Repository], [version], [day month year], [DOI or URL].
Chicago treats a preprint much like an unpublished manuscript, with the repository and identifier:
[Surname], [First name]. "[Title]." Preprint, submitted [date]. [Repository]. [DOI or URL].
IEEE uses its report or online-source pattern with the repository as the source and the identifier attached.
One thing to avoid: "Advance online publication" is not a preprint label. In APA it marks a peer-reviewed article that a journal has published online ahead of assignment to an issue — it stands in place of the volume, issue and page numbers. A preprint has not been accepted by a journal at all, so putting "Advance online publication" on one tells the reader the opposite of the truth. Use the preprint label and the repository.
A Worked Example: When the Preprint Becomes an Article
This is the case the rule exists for. A bioRxiv preprint, posted as v1 on 31 March 2020 under the title "Susceptibility of ferrets, cats, dogs, and different domestic animals to SARS-coronavirus-2" (https://doi.org/10.1101/2020.03.30.015347 — take the full author list from the record itself), was published two months later in Science:
Shi, J., Wen, Z., Zhong, G., Yang, H., Wang, C., Huang, B., … Bu, Z. (2020). Susceptibility of ferrets, cats, dogs, and other domesticated animals to SARS–coronavirus 2. Science, 368(6494), 1016–1020. https://doi.org/10.1126/science.abb7015
Cite the second one. And note what changed between them: the title is not identical — "different domestic animals" in the preprint became "other domesticated animals" in the article, and the hyphenation of the virus name changed. If you had copied the preprint's title into your reference list and pointed it at the journal DOI, you would have produced a reference that matches neither version.
Both records were checked on 22 September 2026: the preprint through the bioRxiv API, which reports its published DOI on the record itself, and the article through Crossref.
Working Papers: Extra Considerations
Working papers carry series numbers, and the number is how a reader finds the document. Always include it.
A real example, verified in Crossref on 22 September 2026:
Song, H., McKenna, R., Chen, A., David, G., & Smith-McLallen, A. (2021). The impact of the non-essential business closure policy on Covid-19 infection rates (NBER Working Paper No. 28374). National Bureau of Economic Research. https://doi.org/10.3386/w28374
The issuing body carries information too. A reader who knows the field reads "NBER Working Paper" differently from an unfamiliar departmental series, so give enough that they can judge it for themselves rather than take your word.
The Version Problem
Preprints get revised, and this creates two distinct problems.
Page and section references move. A passage on page 12 of v1 may be on page 14 of v3, or gone. If you cite a specific location, cite the version you read.
The content itself changes. A claim you cited may have been revised, qualified or dropped. Check the record again before you submit — the servers show the version history, and bioRxiv and medRxiv also show whether a journal version now exists.
Record the version number (v1, v2) or the posting date in your citation, so a reader can get to the text you actually read rather than the one that exists now.
What If the Preprint Gets Published?
Update the citation to the published version before you submit. That is standard practice, and the worked example above shows why: details change.
If you are citing something that changed between the versions, you may cite both — the preprint for the original claim, the article for the revision — and say in your text that this is what you are doing.
If the preprint is withdrawn, remove the citation unless the withdrawal itself is your subject.
Should You Cite Preprints at All?
Reasonable uses: establishing that a question is being worked on; citing very recent findings in a field where publication lags by a year or more; fields whose norms already treat preprints as citable, as much of physics, mathematics and computer science does.
Risky uses: a foundational claim your argument depends on; a field where referees expect every cited source to have been refereed; work likely to be substantially revised.
Where you can, pair a preprint citation with peer-reviewed work that supports the same point. That way a revision to the preprint does not take a plank out of your argument.
Sources
- bioRxiv API record for 10.1101/2020.03.30.015347, including its
publishedDOI field. https://api.biorxiv.org/details/biorxiv/10.1101/2020.03.30.015347 (checked 22 September 2026) - Crossref record for the published article, 10.1126/science.abb7015 — Science 368(6494), 1016–1020, 29 May 2020. (checked 22 September 2026)
- Crossref record for NBER Working Paper No. 28374, 10.3386/w28374 — National Bureau of Economic Research, January 2021. (checked 22 September 2026)
Exact reference formats belong to the style manuals, and the preprint source type has changed between editions; apastyle.apa.org blocks automated requests, so this page gives templates and directs you to your style's own examples page rather than asserting an element order we could not read today.
