Submission policy

Last updated 2026-05-19. Plain English.

Who may submit

Anyone with a verified identity. Independent researchers without institutional backing are welcome on the same footing as people affiliated with universities, research labs, or industry teams. We do not require an institutional email and we do not require an endorser.

Sign in with ORCID (recommended for researchers) or with a did:plc through Bluesky. Either binds the submission to an identifiable real person and persists across the OpenXiv App View's AT-proto record set.

Scope: research, not advocacy

OpenXiv is an academic preprint server. We accept original or review work that uses established scholarly methodology: mathematical proofs, empirical studies with reproducible data, formal analyses, literature reviews with traceable sourcing, structured argumentation grounded in evidence.

We do not accept opinion or advocacy pieces, political manifestos, personal essays, press releases, or marketing material. These restrictions are content-agnostic: a methodologically rigorous study of a sensitive topic (mathematical analysis of conflict demographics; a peer-reviewable survey of any population's outcomes; a legal analysis under any specific framework) is welcome. A persuasive essay on the same topic without academic methodology is not.

We refuse pieces based on academic quality, never on political stance. The line is methodology, not topic.

What we accept

• Preprints in any discipline OpenXiv covers (multidisciplinary; see the category list on /submit).
• LaTeX source — a .tex file or a .zip / .tar.gz bundle containing the .tex, figures, and .bib.
• AI-assisted work, under structured disclosure (see below).
• Revisions of previously-published versions via the version flow on the abstract page.

What we don't accept

• PDF-only uploads (temporarily disabled until the GROBID + Nougat fallback ships).
• Unverified or hallucinated output: hallucinated references, fabricated math, "as-an-AI-language-model" boilerplate left in the body without verification.
• Plagiarism — verbatim or near-verbatim copy of prior work without attribution.
• Material already retracted by another venue for fraud, misconduct, or fabrication.
• Promotional content (vendor whitepapers, marketing collateral styled as research).
• Opinion or advocacy pieces of any political or social stance — see the Scope section above for the criterion.

AI disclosure and rules

Every submission selects one of four disclosure levels:

• None — no AI was used.
• Assistant — used for spell-check, light copy-editing, or other narrow assistance.
• Co-author — used for material drafting, derivations, or structured argument generation, then verified by you.
• Primary — AI produced the bulk of the manuscript; you reviewed and attest to correctness.

The disclosure also asks you to list the models used (e.g. "Claude Sonnet 4.5", "GPT-4o", "DeepSeek V3") and a short note on how each was used. This record is part of your AT-proto submission blob.

The full set of AI-use rules — including the prohibition on listing AI as an author, mandatory citation and numerical verification, and the no-boilerplate rule — is in the code of conduct. A misrepresented disclosure level or a violation of those rules is grounds for refusal.

Refusal — public, specific, revisable

When we refuse a submission, we publish a refusal packet that names the failure mode (e.g. "fabricated citations in §4", "unverified numerical claim in §2.1"), points at the evidence, and recommends a revision. You can resubmit a corrected version. We do not ban authors for refused work.

Versioning

Each version gets a suffix (v1, v2, …). Previous versions stay accessible at their own URLs; new versions inherit the OpenXiv id and DOI of the original. A changelog is required on every version after v1.

Related: Submission terms · Takedown · All policies.