---
title: Linked Data
description: Dereference a public artifact as a CIDOC-CRM JSON-LD graph.
section: protocols
order: 3
updated: 2026-07-11
verified: 2026-07-11
related: [formats/metadata-formats, protocols/oai-pmh, protocols/iiif]
features: [metadata-export]
---

# Linked Data

Every public artifact has a linked-data representation you can fetch directly, no harvesting or authentication involved.

## Request

```
GET https://{your-domain}/{institution}/artifacts/{artifactId}.jsonld
```

This mirrors the artifact's ordinary `.json` representation but returns `application/ld+json`: a [CIDOC-CRM](http://www.cidoc-crm.org/) graph describing the artifact instead of a flat field list. Only public artifacts resolve; anything else returns "not found" without distinguishing missing from non-public, matching the rest of the public site.

## Shape

The response is a flattened JSON-LD document: a top-level `@context` mapping a handful of CURIE prefixes (`crm`, `dcterms`, `skos`, `rdfs`, `owl`) to their namespace IRIs, plus a `@graph` array of nodes. Each node has an `@id` (either a real IRI or a document-scoped blank node id), an `@type`, and predicate values.

## CIDOC-CRM mapping

The artifact itself is an **E22 Human-Made Object**, identified by its public page URL. Its title and identifier attach through **E35 Title** and **E42 Identifier** nodes. Creation details — date, creators, place of production, technique — all hang off a single **E12 Production** event, which is how CIDOC-CRM keeps those facts connected to *one* creation act rather than as independent, disconnected fields. Classified concepts (object type, genre, subject, material, and similar) map to **E55 Type** or **E57 Material** nodes depending on their role. A few Dublin Core Terms literals (`dcterms:title`, `dcterms:description`, `dcterms:type`, `dcterms:rights`) ride alongside the CRM structure for consumers who want the basics without walking the event graph.

## Authority-linked identity

When a creator, place, or classified concept carries a known external authority URI (for example a Getty AAT, Library of Congress, or Wikidata identifier), that URI becomes the node's `@id` directly — rather than a bare label or a `owl:sameAs` pointer to a separate node. Two artifacts (from the same institution or different ones) that cite the same authority-linked term therefore resolve to the *same* node identity, which is what lets linked-data aggregators deduplicate and connect records across collections. A term with no known authority becomes a blank node carrying just its label.

Authority scheme and URI values are always carried through exactly as the institution recorded them — never invented or resolved on export.

## Format

Currently JSON-LD only. There's no Turtle or RDF/XML serialization and no bulk/batch linked-data harvest — this is a single-artifact, per-request representation.
