DOI

DOIs are issued by Registration Agencies (RAs) operating under the International DOI Foundation, the principal RAs for research being Crossref (publishers’ content) and DataCite (research data, software, instruments). A DOI is opaque and case-insensitive, with the prefix ’10.NNNN’ and a suffix chosen by the registrant. The DOI Handle System provides resolution; metadata is registered with the issuing RA and exposed via that RA’s APIs. A DOI commits the registrant to keep the identifier resolvable, even if the underlying object is retracted or moved.

DOI format and structure

Every DOI name has the same two-part shape, defined by ISO 26324 (the DOI syntax standard, currently in its third edition, ISO 26324:2025) and the DOI Handbook maintained by the International DOI Foundation:

<DOI prefix> / <DOI suffix>

the two parts separated by a single forward slash. As a resolvable link, that same DOI name is appended to the DOI resolver base URL:

https://doi.org/<prefix>/<suffix>

Anatomy of a sample DOI

Using the worked example https://doi.org/10.1038/nature14539 (a Nature article):

• Resolver base — https://doi.org/. This is how a DOI is dereferenced as a clickable, persistent link; the bare identifier 10.1038/nature14539 is equally valid on its own, e.g. in a citation string, and resolves the same way once the base is prepended.
• Directory indicator — the leading 10. Every DOI prefix starts with 10; it is the fixed namespace marker that tells the Handle System “this is a DOI name,” distinct from other Handle System namespaces.
• Registrant code — 1038, following the directory indicator after a period. Assigned by the Registration Agency to a specific registrant (here, Springer Nature). Together, 10.1038 is the full DOI prefix.
• Suffix — nature14539, chosen by the registrant. The suffix has no fixed grammar: it can be short and opaque, or carry embedded structure the registrant finds convenient (see below).

A second worked example, this time a data DOI — https://doi.org/10.5281/zenodo.4317001: prefix 10.5281 (registrant code 5281, held by Zenodo/CERN for repository deposits), suffix zenodo.4317001, in which the registrant has chosen to embed the platform name and an internal record number — a convention, not a syntax requirement.

The DOI Handbook’s own canonical illustrative example, 10.1000/182, shows the same two-part shape with the shortest realistic suffix: prefix 10.1000, suffix 182.

Syntax rules

• Prefix characters — numeric digits and the period character only. A period inside the prefix delimits a hierarchical sub-level in the namespace allocation (most registrants use a single flat registrant code, e.g. 10.1038, but a prefix can in principle be further sub-delegated, e.g. 10.1038.1).
• Suffix characters — any printable Unicode character is permitted; there is no ISO-defined length limit on the suffix, the prefix, or the DOI name as a whole. In practice, RAs recommend registrants keep suffixes reasonably short, URL-safe (percent-encoding applies to characters outside the unreserved URI set when the DOI is expressed as part of a URL), and free of embedded semantics that could later become misleading (a page-range suffix stops being accurate if the article is later repaginated, for example).
• Case-insensitivity — DOI names are case-insensitive for resolution purposes. 10.1038/NATURE14539 and 10.1038/nature14539 resolve to the same registered object, though the registrant’s originally-declared casing is what is typically displayed in citations and metadata.
• No embedded meaning is required — unlike an ISBN or ISSN, nothing in a DOI’s syntax encodes the object’s type, publisher identity beyond the registrant code, publication date, or subject matter. A DOI is an opaque identifier by design; any human-readable structure in a suffix is a registrant convention, not a rule enforced by the DOI system itself.

Common formatting mistakes

• Treating the resolver URL and the bare DOI name as different identifiers — they are not. https://doi.org/10.1038/nature14539 and 10.1038/nature14539 identify the same object; only the resolver prefix differs.
• Including a literal DOI: or doi: label as part of a pasted or copied identifier (e.g. DOI: 10.1038/nature14539) — this is a common citation-style convention for display, but the DOI name itself is only the string after the label, and including the label breaks automated parsing/lookups.
• A shortDOI (e.g. via doi.org’s own shortener, or a third-party link shortener) is not itself a registered DOI name and does not follow DOI prefix/suffix syntax — it is a redirect alias to one. Treat it as a convenience link, not as the canonical identifier to cite or archive.
• A missing or malformed directory indicator — a string like 1038/nature14539 (no leading 10.) or 10/1038.nature14539 (slash and period transposed) is not valid DOI syntax and will not resolve.

References

• ISO 26324:2025, Information and documentation — Digital object identifier system (3rd edition).
• International DOI Foundation, ‘DOI Handbook’ — DOI Name Syntax and Numbering sections (doi.org).