Metadata policy

Last updated 2026-05-18. Plain English.

License

Bibliographic metadata for every OpenXiv record is licensed CC0 1.0 (public domain). That means: copy it, reuse it, index it, mirror it. Attribution is appreciated but not required. The metadata covers title, authors, abstract, keywords, categories, DOI, ORCID, publication date, version number.

This is separate from the content license (the PDF / HTML), which authors choose at submit — see the content policy.

Standards we emit

• OAI-PMH 2.0 at /oai-pmh — harvesters can pull `oai_dc` (Dublin Core) records by category or modification window.
• Crossref deposit XML — when a paper's DOI is deposited, the metadata follows Crossref schema 5.3.1.
• HighWire Press citation_* meta tags on each abstract page (`citation_title`, `citation_author`, `citation_publication_date`, `citation_doi`, `citation_pdf_url`, `citation_arxiv_id`, `citation_journal_title`, `citation_issn`). This is what Google Scholar parses.
• JSON-LD ScholarlyArticle in the page head with author Person + affiliation Organization nodes and an `isPartOf` link to the OpenXiv ISSN.
• Open Graph + Twitter cards for social previews.
• Atom feed at /feed.atom.

Languages

OpenXiv currently emits metadata in English (citation_language=en, inLanguage: 'en'). Multi-language metadata is on the roadmap; submissions in other languages may declare their language in the abstract, and the bibliographic surface will pick that up when the schema lands.

Updates and corrections

Bibliographic data is immutable after publish: title, author list, abstract, DOI — none of these change on the original record. Corrections take one of two forms:

• Versioning — submit a new version (v2, v3, …) with the corrected content. Each version has its own record; the original stays accessible.
• Crossref correction — when the corrected version is deposited, Crossref's update record links it back to the original so downstream indexers can follow the chain.

The same correction also propagates to ORCID for every author with an ORCID iD recorded — their work list at orcid.org auto-updates from the deposit.

Identifiers

Every published paper carries:

• An OpenXiv id like cs.AI.2026.00001 — short, sortable, year + subject scoped.
• A DOI when Crossref membership credentials are provisioned. DOIs are opaque (10.{prefix}/openxiv.{id}) and immutable.
• An AT-proto URI on the author's Personal Data Server, so the record federates with Bluesky.

Related: Content policy · Preservation · All policies.