Digital Object Identifier (DOI)

The core function of a DOI is to provide a permanent link that never changes. When a publisher moves content to a new server or changes their URL structure, they update the DOI record in the central resolver. All existing DOI links automatically redirect to the new location, eliminating broken links and link rot in scholarly and professional literature.

Every DOI is registered with a rich set of bibliographic metadata defined by the DataCite Metadata Schema or Crossref schema. This includes:

• Creator names with ORCID identifiers
• Publication date and publisher information
• Resource type (journal article, dataset, software, etc.)
• Funding references and licensing information This structured metadata enables machine-readability and automated citation formatting across thousands of systems.

The DOI system operates on the Handle System, a distributed computer network maintained by the DONA Foundation. This infrastructure ensures global resolvability independent of any single organization. DOIs are interoperable with:

• ORCID for author disambiguation
• ROR for institutional identification
• Funder Registry for grant tracking
• Crossref and DataCite for citation linking This creates a connected graph of scholarly entities.

A DOI can be assigned to specific versions of a digital object, not just the abstract work. This enables precise citation of:

• Preprints versus final published versions
• Dataset revisions with distinct version numbers
• Software releases at specific commits Each version receives its own DOI, while a parent DOI may link to all versions, supporting reproducible research and accurate provenance tracking.

DOIs serve as the backbone of citation linking in scholarly communication. When a paper cites another work by its DOI, the citation graph becomes machine-traversable. This powers:

• Citation count tracking for impact metrics
• Reference resolution to verify citation accuracy
• Attribution chains that credit original authors
• Plagiarism detection systems that match text to DOI-registered sources A DOI citation is a verifiable, auditable link in the knowledge graph.

The computational task of mapping a textual citation string to a specific bibliographic entity in a knowledge base. When a language model cites a DOI, reference resolution involves querying the DOI resolver (doi.org) to retrieve structured metadata—author, title, journal, publication date—and the current source lineage. This ensures the model's citation points to the exact document intended, not a similarly titled work.

The phenomenon where a DOI link becomes non-functional or the underlying content drifts over time. While DOIs are designed to be persistent identifiers, decay occurs when publishers fail to update the DOI record's target URL or when the registration agency's infrastructure is disrupted. In generative AI contexts, this undermines citation integrity, as a model may cite a DOI that no longer resolves to the original source material.

A network model where DOIs serve as the primary node identifiers, and directed edges represent citation relationships between works. These graphs power source authority scores by analyzing patterns such as:

• Citation frequency: How often a work is referenced
• Co-citation clusters: Groups of works frequently cited together
• Bibliographic coupling: Works that share common references Generative AI systems use citation graphs to prioritize high-authority sources during source grounding.

The process of cryptographically validating that a DOI's associated provenance ledger is authentic and untampered. This involves checking digital signatures from the registration agency and verifying the hash chain linking each update to the DOI record. For AI-generated content with citations, provenance verification assures users that the cited DOI genuinely represents the work claimed and hasn't been spoofed or manipulated.

Structured data markup formats, such as Schema.org's ScholarlyArticle type, that embed machine-readable DOI information directly into web pages. Key properties include:

• identifier: The DOI string
• citation: Links to referenced works by their DOIs
• author and datePublished: Core bibliographic metadata This markup enables AI crawlers to extract precise citation data without relying on brittle screen scraping, improving citation transparency.