How to cite a bioRxiv or arXiv preprint (APA 7, Harvard)

Preprints have been mainstream in academia since COVID. Anyone researching a current topic in 2026 — in medicine, economics, computer science, psychology — will hit arXiv, bioRxiv or SSRN results all over Google Scholar. Some of those studies later land in Nature. Others quietly disappear. For a bachelor’s or master’s thesis that means: you have to decide whether to cite an unreviewed early version, and if so, exactly how. This guide sorts it out.

What a preprint is — and isn’t

A preprint is the manuscript version of a research paper that the authors upload to a public server before the paper has gone through peer review at a journal. It is the raw draft, with a DOI and a citable form, but without external quality control.

Important distinctions:

• A preprint is not a “working paper” from a research department (similar idea, but typically not on a central server).
• It is not the “Author Accepted Manuscript” (AAM) that publishers allow after peer review — that’s a postprint.
• It is not a predatory journal article. Predatory journals pretend they have peer review. Preprint servers explicitly say they don’t.

The major preprint servers

Depending on your field you’ll run into different servers. The five large ones:

Alongside these you’ll find PsyArXiv (psychology), ChemRxiv (chemistry), EarthArXiv and several discipline-specific platforms. All issue DOIs, all carry version numbers (v1, v2 …), and all display a prominent disclaimer: “This article is a preprint and has not been peer-reviewed.”

Are preprints even allowed?

First rule: ask your department. Some faculties forbid preprints in graded theses, others allow them if you flag them explicitly. That policy rarely shows up in the official handbook, but it shows up regularly in how examiners grade.

When the policy is silent, a useful rule of thumb: preprints are acceptable when (a) no peer-reviewed version exists, (b) you flag them as preprints, and (c) the claim you’re supporting describes the state of research rather than an established fact.

Sensible uses:

• Recent methodological developments not yet in journals (typical in ML/AI).
• Data from ongoing crises or fast-moving research fields.
• Conference submissions shared as preprints before they appear in proceedings.

Risky to unsuitable:

• Load-bearing claims your whole argument rests on.
• Clinical recommendations or efficacy claims (medRxiv carries its own warnings about exactly this).
• Sources whose methodology you can’t evaluate yourself.

Before citing: check whether it has already been published

The most common beginner mistake: cite a 2023 preprint even though the final version came out in a journal in 2024. You forfeit the peer-reviewed source and cite the weaker version.

How to find the final version:

1. Look for “Published in” on the preprint page. arXiv, bioRxiv and medRxiv link the final version automatically once it appears.
2. Search the DOI in Google Scholar. Scholar usually merges preprint and journal versions.
3. Check the title in Crossref or Semantic Scholar. Crossref gives you the full DOI history.

If a peer-reviewed version exists, cite that — not the preprint. Full stop.

The version trap: v1, v2, v3

Preprints get revised. arXiv v1 from 2024 may differ substantially from v3 in 2026 — different data, different conclusions, sometimes different authors. If you cite the preprint, you must specify the exact version, otherwise you’re citing something that no longer exists in that form.

In APA 7 you handle this via the version date:

Müller, A., & Weber, T. (2024). Scaling language models on embedded devices (Version 2) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2401.12345

If you use Zotero: importing from arXiv/bioRxiv often drops the version date from the “Date” field. Add it manually.

Citing correctly in the four main styles

Worked example: a 2025 bioRxiv preprint, author Smith, title “Gut microbiome dynamics under intermittent fasting”, DOI 10.1101/2025.03.15.123456.

APA 7

Smith, J. (2025). Gut microbiome dynamics under intermittent fasting [Preprint]. bioRxiv. https://doi.org/10.1101/2025.03.15.123456

The [Preprint] in square brackets is the official APA 7 marker for unreviewed manuscripts. In-text: (Smith, 2025).

Harvard

Smith, J. (2025) Gut microbiome dynamics under intermittent fasting. bioRxiv [Preprint]. Available at: https://doi.org/10.1101/2025.03.15.123456 (Accessed: 29 June 2026).

An access date is standard in Harvard for online sources and especially useful for preprints because versions change.

Chicago (author-date)

Smith, Jane. 2025. “Gut Microbiome Dynamics under Intermittent Fasting.” Preprint, submitted March 15. bioRxiv. https://doi.org/10.1101/2025.03.15.123456.

In-text: (Smith 2025).

MLA 9

Smith, Jane. “Gut Microbiome Dynamics under Intermittent Fasting.” bioRxiv, 15 Mar. 2025, https://doi.org/10.1101/2025.03.15.123456. Preprint.

The trailing “Preprint.” is an optional supplemental element MLA 9 explicitly provides for.

In all four styles the preprint marker is mandatory. If your style guide PDF is silent on it, use the examples above as a template — no examiner will object.

DOIs from the preprint server, never from a search engine

The DOIs you see on Google Scholar or ResearchGate can point to a commercial mirror, not the original. Always pull the DOI directly from the preprint server (arxiv.org, biorxiv.org etc.). On arXiv it starts with 10.48550/arXiv., on bioRxiv and medRxiv with 10.1101/, on SSRN with 10.2139/ssrn..

Withdrawn and disputed preprints

Unlike journal articles, where Retraction Watch picks up corrections, preprints tend to be “withdrawn” or get a warning banner. During COVID there were high-profile cases (hydroxychloroquine studies on medRxiv) that were cited thousands of times before they were pulled.

Before citing: go to the preprint page and check whether the header shows a banner (“Withdrawn”, “Cited concern”, “Replaced by”). If it does: don’t cite it, unless you’re specifically discussing that case.

Pre-submission checklist

• Is there a peer-reviewed version? → Cite that, not the preprint.
• Is the preprint flagged as such in your reference list ([Preprint] or equivalent)?
• Does the citation include the version number when you mean a specific version?
• DOI taken directly from the server, not from a mirror?
• No withdrawal banner on the preprint page?
• Does your institution allow preprints at all? (When in doubt, send a one-line email.)

A preprint tells you what is being researched, not what is settled

The value of preprints is speed. They show you what’s currently being worked on, before the publishing process delays things by 12 to 18 months. The price is the missing external check. If you make that transparent in your argument (“a not-yet-peer-reviewed preprint indicates …”), you’re on safe ground.

What a preprint guarantees no more than a journal article does: that the claim you attribute to the source actually appears there, and that your entry still points at the version you read.

The Quick Check looks up whether the preprint appears in the registers and whether DOI, title and year hang together. It does not tell you whether a journal version already exists, and it does not read the preprint full text against your sentences. A prefilled example shows how an entry with a wrong or dead DOI stands out. Check a sample citation →

The analysis with an upload does the content step: Acurio matches every citation in your thesis against the preprint PDF you upload. Does the claim match the passage? Is the page number right? That is the check peer review does not provide when you cite a preprint.

Preprints are a sharp tool: fast, current, openly accessible — and unreviewed. Flag them carefully, prefer the final version when one exists, carry the version number with you, and they’re no problem in a bachelor’s or master’s thesis. Treat them like ordinary journal articles, and that’s what stands out.