arXiv (pronounced "archive") is the world's definitive STEM preprint server. Founded in August 1991 by physicist Paul Ginsparg at Los Alamos National Laboratory, it moved to Cornell University in 2001 and, as of July 2026, operates as an independent non-profit organization funded by member institutions, Simons Foundation International, and Schmidt Sciences. It hosts more than 3 million scholarly articles and receives roughly 24,000 new submissions every month.

arXiv is not a journal. It performs no peer review, charges no fees, and makes every paper freely readable by anyone. What it provides is speed and reach: your results become citable and publicly timestamped within a day or two of submission, months or years before a journal version appears. In many fields - high-energy physics, machine learning, mathematics - posting to arXiv is simply how research is announced.

arXiv accepts submissions in eight fields, organized into ~150 moderated categories:

If your work is biomedical or social-science research, a sibling server (bioRxiv, medRxiv, SSRN, PsyArXiv, SocArXiv) is usually the better home.

1. Register an account at arxiv.org with an academic email where possible.
2. Get endorsed if required. Since 2004, first-time submitters to a category may need an endorsement from an established arXiv author. Researchers at recognized institutions are typically endorsed automatically.
3. Prepare your files. arXiv strongly prefers (La)TeX / PDFLaTeX source - it compiles your source itself. Plain PDF is accepted where source is unavailable; DVI, PostScript, and scanned documents are rejected.
4. Choose a license. Options: the arXiv perpetual non-exclusive license (most conservative - check your target journal's policy), CC BY, CC BY-SA, CC BY-NC-SA, CC BY-NC-ND, or CC0. The choice is irrevocable, so verify funder and journal requirements first.
5. Submit and wait for moderation. Volunteer subject moderators screen for scope, scholarly relevance, and policy compliance - typically clearing within 1-2 business days.
6. Announcement. Your paper receives an arXiv ID, a DataCite DOI, and appears in the daily listings and RSS feeds for its categories.

arXiv papers are not peer reviewed. Instead, submissions pass through automated checks plus human moderation by volunteer subject experts, who verify that a paper is topical, scholarly in form, and not plagiarized or previously published spam. Moderators may reclassify a paper into a better-fitting category or reject it entirely; they do not evaluate correctness or significance. Treat an arXiv preprint as a claim awaiting scrutiny - the community and eventual journal review provide the quality filter.

Every paper gets a permanent identifier in the form arXiv:YYMM.NNNNN (e.g., arXiv:2501.00001), with an optional version suffix (v1, v2, ...). Since April 2007 the ID is category-independent; older papers use the archive/YYMMNN form (e.g., astro-ph/0107001). Without a version suffix, the ID always resolves to the latest version.

All arXiv papers are also automatically assigned a DataCite DOI of the form 10.48550/arXiv.2501.00001, which points to the abstract page and always resolves to the newest version - new versions update the metadata rather than minting a new DOI.

To cite a preprint, arXiv recommends including the specific version, category, and date, mirroring the PDF stamp: arXiv:0706.0001v1 [q-bio.CB] 1 Jun 2007. Most citation styles (APA, IEEE, Vancouver) now have explicit arXiv/preprint formats; SciSpace's citation generator supports them all. If a journal version exists, cite that instead and link the arXiv ID as supplementary access.

Posting to arXiv does not count as prior publication for the vast majority of physics, math, CS, and statistics journals - most explicitly welcome preprints, and some (notably in machine learning) effectively require them. The standard workflow is: post the preprint, submit to a journal or conference, then update the arXiv record with the accepted version (check the journal's policy on which version you may post) and add the journal reference and publisher DOI to the arXiv metadata. A small number of medical and society journals still have restrictive preprint policies, so check the specific journal's guidance - Sherpa Romeo and the journal's author pages are authoritative.

Revisions are cheap: uploading a corrected v2 takes minutes, and prior versions remain permanently accessible, giving you a transparent public record of the paper's evolution.

arXiv charges nothing - no submission fees, no APCs, no reader paywalls. Operations are funded by a membership model (universities, labs, libraries, and consortia), together with grants from Simons Foundation International and Schmidt Sciences. This is worth appreciating: the entire service, at 24,000 papers a month, runs on collaborative funding rather than author fees.

Use arXiv when you work in its eight fields and want fast, free, citable dissemination with a timestamp that establishes priority. Skip it (in favor of a sibling server or direct journal submission) when your target journal prohibits preprints, when your work involves clinical findings better suited to medRxiv's specialized screening, or when your field's norms route through SSRN or OSF-hosted servers. When in doubt, search arXiv for recent papers from your subfield - if your peers post there, you should too.