SiteNexis is an AI Retrieval and Machine Trust Intelligence platform that models how AI systems — including ChatGPT, Gemini, Perplexity, and Claude — retrieve, interpret, trust, and recommend web content. It runs 16 intelligence agents across four layers and produces a 12-dimensional audit report with actionable recommendations.

How does SiteNexis improve AI visibility?

SiteNexis analyzes entity clarity, schema completeness, chunk extractability, semantic trust signals, and retrieval simulation across six stages. Each finding maps to a named issue with specific recommendations. Improvements to entity consistency, schema alignment, and content structure directly raise the probability of being retrieved and recommended by AI systems.

Which AI systems does SiteNexis analyze?

SiteNexis models retrieval behavior across Google AI Overviews, ChatGPT browsing and RAG modes, Perplexity, Gemini, Claude web search, voice assistants, and autonomous agents. Provider scoring weights are configurable and all scores are labeled as probabilistic estimates.

What is the Machine Trust Score?

The Machine Trust Score is a 0–100 composite that models AI system confidence in a website as a reliable source. It combines entity credibility consistency (30%), schema trust alignment (20%), external validation depth (25%), contradiction absence (15%), and trust degradation resistance (10%).

What is the Intelligence Report?

The Intelligence Report is a Grok-style editorial narrative generated for every audit. It synthesizes all twelve scores and the top issues into six named sections: a Signal Summary, SEO Posture, AI Visibility standing, Machine Trust analysis, Entity Health, and a Strategic Verdict with a composite X.X/10 score. Unlike dashboards that surface raw numbers, the Intelligence Report reads like a professional analyst's assessment — prose with precision.

What is the Decision Roadmap?

The Decision Roadmap is the output of the Decision Orchestrator — a v4 system that transforms a ranked issue list into an optimally sequenced action plan using Sequencing Value Score (SVS). It identifies the single best action to start with, groups subsequent actions into This Week, This Sprint, and Backlog bands, and calculates the projected score impact of completing each band. It also identifies parallel tracks for team-based execution.

SiteNexis has a free tier with 1 full AI visibility audit per month covering Layers 1–3. Layer 4 Machine Trust Intelligence — including Retrieval Simulation, Machine Trust Score, Temporal Authority, Recommendation Surface Mapping, and Synthetic Entity Detection — is available on Pro and Agency plans.

Back to Blog

Perception vs Fact Layer 6 min readJun 16, 2026

Part 5 of 10Perception vs Fact Layer series

Perception Graph Relationship Types: isA, partOf, contradicts, supports

The AI Perception Graph uses typed edges to represent different kinds of relationships between entities and topics. The relationship type determines how the AI system uses the connection in retrieval and citation.

Not all relationships are the same. "SiteNexis is a software platform" is a classification relationship. "Machine Trust Score is part of the SiteNexis scoring system" is a composition relationship. "This page contradicts that page on entity founding date" is a contradiction relationship. The AI Perception Graph uses typed edges to distinguish these fundamentally different relationship kinds — because the type of relationship determines how the AI uses the connection.

The Eight Perception Edge Types

isA — classification: "X is a type of Y." Used for entity type hierarchy.
partOf — composition: "X is a component of Y." Used to link sub-features to primary entities.
relatedTo — semantic proximity: "X is related to Y." Weakest relationship type — used when a specific type cannot be determined.
contradicts — contradiction: "X and Y make conflicting claims about the same fact." Negative trust signal.
supports — corroboration: "X corroborates or reinforces a claim made by Y." Positive trust signal.
authorOf — attribution: "Person X authored content Y." Used for authorship trust signals.
locatedIn — geographic: "Entity X is located in place Y." Used for LocalBusiness entity disambiguation.
offers — service/product: "Organisation X offers service/product Y." Used for commercial entity modeling.

How Relationship Types Affect Retrieval

The contradicts relationship type has the most significant downstream effect. When the Perception Graph contains a contradicts edge between two perception nodes, it is a signal that AI citation eligibility should be reduced for the content involved. The AI system cannot cite both versions of a contradictory claim. The contradiction degrades the reliability of the entire content cluster around the contradicting entities, not just the specific contradicting claims.

How Relationship Types Are Inferred

The AI Perception Graph construction uses a combination of schema markup analysis (Organization, Product, Person, Place schemas map directly to entity types and relationships), entity co-occurrence analysis (entities that appear in the same chunk and the context of their co-occurrence), and Claude API semantic analysis for the top pages by PageRank. The inferred relationship type carries a confidence score that reflects how clear the evidence for that specific relationship type was.

Perception vs Fact Layer

Fact Graph vs Perception Graph: The Core Distinction One-Directional Derivation: Why the Fact Graph Never Reads from Perception Topical Clusters in the AI Perception Graph

Perception vs Fact Layer — Part 5 of 10

PreviousEntity Disambiguation in the Perception Graph NextCitation Readiness vs Retrieval Readiness: Two Different Goals

Tags: perception graph relationship types graph edges Entity Relationships

All articles Run free audit