An ARK (Archival Resource Key) is a persistent identifier scheme designed for information objects of any type — a page image, a dataset, a finding aid, a physical object record, a person, an organization — that is resolvable through a distributed network of cooperating resolvers rather than a single central registry. Where a DOI is issued through a paid registration agency (Crossref or DataCite) built for scholarly publishing, and an ORCID iD and ROR ID identify people and organizations respectively, an ARK is free to mint, has no publishing-industry framing, and is the identifier most likely to already be embedded in a national library’s, archive’s, or museum’s holdings. This guide explains what ARKs are, who governs the specification today, how the syntax and resolution model differ from DOI, and where research administrators and CRIS managers actually encounter them.
What an ARK looks like and what it identifies
An ARK takes the form ark:/NAAN/Name[Qualifier], where NAAN is the Name Assigning Authority Number (usually a five-digit number, sometimes a longer “shoulder”) assigned to the organization minting the identifier, and Name is the local, organization-assigned identifier string. A real example from the Bibliothèque nationale de France is ark:/12148/btv1b8449691v, identifying a specific manuscript; a California Digital Library example is ark:/13030/kt9c6035s2. Both resolve through cooperating resolvers, most commonly the ARK Alliance’s own N2T (Name-to-Thing) global resolver at n2t.net, without requiring the minting organization to run resolver infrastructure of its own.
Two features distinguish ARK’s design from most other PID schemes:
- The inflection rule. Appending a single question mark to a resolvable ARK URL (e.g.
https://n2t.net/ark:/12148/btv1b8449691v?) is meant to return a brief metadata record about the object rather than the object itself; appending two question marks is meant to return a persistence policy commitment statement from the object’s steward — a machine- and human-readable statement of how long, and under what conditions, the identifier is expected to keep resolving. See the CASRAI dictionary entry on ARK inflection rules. - Structural openness for qualifiers and variants. The optional Qualifier portion of the syntax lets an ARK address a sub-part or version of an object (a specific page of a digitized manuscript, for instance) without minting an entirely separate top-level identifier for it.
CASRAI’s dictionary entry for the base concept is at ARK; this guide goes further into governance, the DOI comparison, and practical use.
Who maintains the ARK specification
ARK was designed by John Kunze at the California Digital Library (CDL) starting in 2001 and documented as an IETF Internet-Draft (not a ratified RFC) across several revisions. For roughly two decades CDL operated the reference resolver and the NAAN registry as a practical convenience for the adopting community, but stewardship has since moved to the ARK Alliance (arks.org) — a community-governed body, supported in part by LYRASIS, that now runs the working groups responsible for the specification, the public NAAN registry, and the N2T resolver, explicitly to transition that infrastructure “from CDL to a community supported and managed activity.” Any stable memory organization — a library, archive, museum, government agency, or research institution — can request a NAAN through the ARK Alliance at no cost and begin minting ARKs immediately; no publishing intermediary, application fee, or per-identifier charge is involved. According to the ARK Alliance’s own published overview, well over 10 billion ARKs have been minted globally, overwhelmingly by libraries, archives, and government memory institutions rather than journal or dataset publishers.
This governance model is the single biggest practical difference from DOI, and it is worth being precise about rather than treating “PID” as one undifferentiated category. See CASRAI’s persistent identifier (PID) entry for the umbrella concept, and the cluster overview at Persistent Identifiers & Research Information Systems for how ARK, DOI, ORCID, ROR, and RAiD relate as a family.
ARK vs. DOI: what actually differs
Both are resolvable, actionable identifiers intended to persist independent of an object’s current location — the surface-level similarity that makes them easy to lump together. The differences that matter operationally:
| Dimension | ARK | DOI |
|---|---|---|
| Registration authority | None required centrally; any organization can request a free NAAN and self-mint indefinitely | Must be minted through an accredited registration agency — Crossref (scholarly literature) or DataCite (data, software, other outputs) are the two research-sector agencies; see CASRAI’s guide to DOI registration agencies |
| Cost | Free to obtain a NAAN and mint ARKs; no per-identifier fee | Registration agencies charge membership and/or per-DOI fees, ultimately funding a centralized metadata infrastructure |
| Metadata requirements | Minimal and locally defined; no mandatory shared metadata schema across all ARK-minting organizations | Registration agencies enforce a required minimum metadata schema (Crossref and DataCite each publish their own) submitted at registration |
| Typical minting community | Libraries, national and university archives, museums, government memory institutions | Scholarly publishers, journals, data repositories, software archives |
| Resolution infrastructure | Distributed; the N2T global resolver plus any organization’s own resolver | Centralized through doi.org (backed by the Handle System) |
| Built-in metadata retrieval | The inflection rule (appending ? / ??) is part of the syntax itself |
Metadata retrieval is a separate API call against the registration agency, not part of the identifier syntax |
Neither model is strictly “better” — they were built for different institutional realities. DOI’s centralized, fee-funded, metadata-enforced model suits scholarly publishing, where consistent, machine-actionable citation metadata across thousands of publishers is the whole point (see how to find, look up, and resolve a DOI). ARK’s free, low-barrier, locally-governed model suits memory institutions that need to mint identifiers at very large scale — for individual digitized pages, archival items, or catalog records — without a per-object budget line for identifier fees.
Where ARKs are actually used
Because ARK’s adopting community is memory institutions rather than journal publishers, the practical use cases a research administrator or CRIS manager is likely to encounter differ from DOI’s:
- National and university library digitization programs. The Bibliothèque nationale de France (BnF), the California Digital Library and its constituent University of California libraries, and numerous other national libraries assign ARKs to digitized manuscripts, photographs, and archival items at scale — often to hundreds of thousands or millions of individual page images within a single collection.
- Web archiving. The Internet Archive uses ARKs extensively for identifying archived web content and other holdings.
- Archival finding aids and special collections. Institutional and consortial archives use ARKs to give persistent, citable identifiers to finding aids, series, and individual archival items described in systems like ArchivesSpace, independent of any future migration of the underlying catalog platform.
- Institutional repositories with heterogeneous, non-scholarly-publication content. Where a repository holds material that doesn’t fit a DOI registration agency’s scope well — physical object records, oral history recordings, locally digitized ephemera — ARK’s low barrier to entry and lack of a required metadata schema make it a practical fit. See CASRAI’s comparison of research repositories vs. archives for how that distinction plays out operationally.
- Cataloging systems that need identifiers below the item level. The qualifier syntax lets an institution mint a coherent, addressable identifier for a specific page, folder, or component of a larger described object without a separate top-level registration for each.
ARKs appear far less often in the scholarly-article and dataset-citation contexts that dominate CASRAI’s DOI content, and an ARK is not a substitute for a DOI where a funder, journal, or repository policy specifically requires one — NIH, NSF, and most journals that mandate a data-citation identifier expect a DOI (typically via DataCite), not an ARK. The two schemes are complementary rather than competing across most of a research institution’s identifier landscape.
Practical guidance for research administrators and CRIS managers
- Don’t assume every persistent-looking identifier from a partner archive is a DOI. If a collaborating national library, museum, or special-collections partner cites material by an
ark:/...string, that is a legitimate, resolvable identifier — test it by pasting it afterhttps://n2t.net/rather than assuming it’s malformed or non-standard. - If your institution runs (or is considering) a digital archive, ARK is worth evaluating alongside a Handle or DOI. The relevant questions are cost (ARK’s NAAN registration is free; DOI registration agencies charge fees), required metadata rigor, and whether the content is closer to “scholarly output needing citation” (favors DOI) or “archival holding needing a stable, low-cost identifier at very large scale” (favors ARK).
- Crosswalking ARKs into a CRIS. Where an institution’s research outputs reference archival source material identified by ARK (a digitized primary source cited in a publication, for example), that ARK can simply be recorded as an external identifier field in the same way an institution would record a DOI, ORCID iD, or ROR ID — see CASRAI’s identifier crosswalk guide for the general mechanics of linking identifier schemes in real metadata records, and persistent identifier federation across research infrastructure for how multiple PID schemes coexist within one institution’s systems.
Frequently asked questions
Is an ARK the same thing as a DOI?
No. Both are persistent, resolvable identifiers, but a DOI must be registered through an accredited registration agency (Crossref or DataCite in the research context) for a fee and under a required metadata schema, while an ARK can be minted for free by any organization that has obtained a NAAN, with no centrally mandated metadata schema.
Who governs the ARK specification today?
The ARK Alliance (arks.org), a community-supported body that has taken over stewardship of the specification, the public NAAN registry, and the N2T global resolver from the California Digital Library, which originally designed and hosted ARK starting in 2001.
Does minting an ARK cost anything?
No. An organization requests a NAAN (Name Assigning Authority Number) from the ARK Alliance at no cost, and can then mint an unlimited number of ARKs under that NAAN without a per-identifier fee.
Where will I actually encounter ARKs in a research context?
Most commonly when working with digitized archival material, finding aids, or web-archived content from a national library, university special collections, or the Internet Archive — rather than in journal-article or dataset-DOI contexts, where DOI remains the dominant identifier.
Can an object have both an ARK and a DOI?
Yes. There is no rule preventing dual identification, and it is not unusual for a digitized item to carry both an institution-assigned ARK (for the archival holding) and, separately, a DOI if it is also formally cited as a dataset or published output.
