The Document Graph is the medium-fuzziness knowledge representation in the Algorithmic Trinity, containing annotated URLs with ranked authority edges and anchor text as labeled relationship descriptors, updating at intervals of days to weeks.

Jason Barnard defines the Document Graph as the knowledge representation where Credibility lives. This is the search engine's index - not a simple list of pages but a rich graph of annotated URLs connected by ranked authority edges, where anchor text serves as labeled relationship descriptors. Fuzziness is medium: edges are scored and ranked rather than binary (Entity Graph) or probabilistic (Concept Graph). The Document Graph updates faster than the Entity Graph (days to weeks rather than weeks to months) and provides the ranked corroboration that transforms Entity Graph facts into trusted claims. When an Assistive Engine grounds a response by searching the web, it is consulting the Document Graph to verify and corroborate what the Entity Graph asserts. This is why third-party corroboration - independent sources confirming entity claims - operates primarily at the Document Graph level.

The search engine index representation - medium fuzziness, ranked authority edges, anchor text as labeled relationships. Where Credibility lives. Updates in days-weeks. Provides the ranked corroboration that transforms Entity Graph facts into trusted claims. The layer Assistive Engines consult during grounding.

Jason Barnard positions Document Graph as the C (Credibility) layer of the Three Graphs. Medium fuzziness, ranked edges, anchor text labels. The corroboration layer consulted during grounding.

┌─────────────────────────────────────────────────────────────┐
│                      DOCUMENT GRAPH                         │
│               "The Investigative Journalist"                │
│                  Fuzziness: MEDIUM ●●○                      │
└─────────────────────────────────────────────────────────────┘

┌───────────────────────────────────────────────────────────┐
│                                                           │
│    ┌──────────────┐                                       │
│    │ kalicube.com │◀────┐                                 │
│    │  (PR: 45)    │     │ citation                        │
│    └──────┬───────┘     │                                 │
│           │link         │                                 │
│           ▼             │                                 │
│    ┌──────────────┐  ┌──────────────┐                     │
│    │ /about-jason │  │searchengine  │                     │
│    │  (PR: 38)    │  │  land.com    │                     │
│    └──────────────┘  │  (PR: 72)    │                     │
│                      └──────────────┘                     │
│                                                           │
│    NODES: Documents (URLs, web pages)                    │
│    EDGES: Links + Topical annotations                    │
│    WEIGHTS: Authority (PageRank) + Relevance scores      │
│                                                           │
│    Scored/ranked - not binary, not probabilistic         │
└───────────────────────────────────────────────────────────┘

OUTCOME: AI CITES you
         Build SECOND (after Entity, before Concept)