The AI Perception Graph: Modeling the Machine View of Your Domain
The AI Perception Graph is not a link graph. It is the semantic representation an AI system builds of your website — nodes for entities, topics, and claims; edges for typed relationships between them. SiteNexis builds and visualizes this graph from your crawled content.
When you navigate your website, you follow links between pages. When an AI system processes your website, it builds a semantic graph of the entities and relationships it encounters. This is the AI Perception Graph: the cognitive model the AI system would build from your content. It exists whether you model it or not. SiteNexis makes it visible so you can see exactly what the AI understands about your domain — and where its understanding is incomplete, contradictory, or confused.
Nodes in the Perception Graph
The Perception Graph has four node types. Entity nodes represent named real-world objects: organisations, people, products, locations. Topic nodes represent semantic concepts that appear repeatedly across pages but are not specific named entities — "retrieval quality," "trust signals," "entity disambiguation." Claim nodes represent specific factual assertions extracted from content: verifiable, attributable statements. Page nodes represent individual URLs as anchor points for evidence attribution. Each node carries a confidence score and a citation readiness score.
Typed Edges and What They Reveal
Edges in the Perception Graph carry typed relationships: isA (X is a type of Y), partOf (X is a component of Y), relatedTo (X and Y are semantically connected), contradicts (X and Y make conflicting claims), supports (X provides evidence for Y), authorOf (person entity X authored content Y), locatedIn (X operates in location Y), offers (organisation X provides service Y). The contradiction edges are the most diagnostically valuable: they show where the AI's model of your domain contains internal inconsistencies.
▲Contradiction edges are the most dangerous finding in a Perception Graph audit. If the AI system's model contains a contradiction edge between your primary entity and one of its claimed attributes, that entity node will receive a reduced confidence score — degrading citation readiness across every page that references the entity.
Graph Analysis Outputs
From the Perception Graph, SiteNexis derives several higher-order outputs. Topical authority clusters show which concept groups are well-supported by multiple pages and which are thin. Trust pathways show which entities have the strongest chain of supporting evidence. Citation pathways identify the nodes most likely to be cited by AI systems — those with high confidence, high evidence density, and no contradiction edges. Contextual gaps show entity relationships that are implied but not explicitly stated in the content.
Visualization and Interaction
The Perception Graph is visualized using react-force-graph — a canvas-based force-directed graph renderer. Node size maps to confidence score. Node color maps to node type. Edge thickness maps to relationship strength. Edge color maps to relationship type. The interactive visualization lets you click any node to see which pages it is supported by, what its citation readiness score is, and what contradiction or gap signals affect it. It is the fastest way to find AI visibility problems that traditional audit tools cannot surface.