Machine Trust Score: How AI Systems Form and Lose Confidence in Your Domain
Trust in AI systems is not a single signal. It is a composite formed from cross-source validation, entity consistency, internal coherence, and the absence of contradictory claims. SiteNexis models all five dimensions.
AI systems do not trust sources the way humans do — by reading about a brand's reputation and forming an opinion. They derive trust from measurable consistency signals: does the entity data agree across multiple pages? Does the schema markup describe content that actually exists on the page? Do external sources confirm what the site claims about itself? When these signals conflict, trust degrades. When they align and are independently verified, trust strengthens. The Machine Trust Score models this five-signal framework across your entire domain.
The Five Trust Dimensions
- Entity Credibility Consistency (30%) — entity attributes consistent across all pages, schema, body text, and metadata
- Schema Trust Alignment (20%) — schema markup accurately describes page content without over-claiming
- External Validation Depth (25%) — sameAs links resolving to authoritative external sources confirming entity claims
- Contradiction Absence (15%) — no conflicting factual claims detected across pages via semantic analysis
- Trust Degradation Resistance (10%) — no signals of trust damage: schema removal, attribute changes, lost validation sources
Entity Credibility Consistency
This dimension scores whether your primary entity — typically your organisation or brand — is described consistently across every relevant page. We check four attributes: name (must match exactly across schema, H1, and body text), foundingDate (must be consistent or absent), description (core description must not contradict itself across pages), and category/type (must consistently match the same schema type). A founding year stated as 2018 on the About page but 2019 in a press release creates a credibility consistency failure.
Schema Trust Alignment
Schema markup that describes more than the page contains is a trust risk, not a trust signal. We check: every schema attribute must be verifiable from the page's body text. A Product schema claiming an aggregate rating of 4.9 on a page with no visible reviews is a schema manipulation signal. An author schema attributing content to a person not mentioned in the body text creates a trust misalignment. Schema trust alignment is fully programmatic — no AI API required.
▲Schema over-claiming is one of the most common trust degradation patterns in our audits. Adding aggregate rating schema to pages without review content, or publishing author schema for entities that do not contribute to the page, creates measurable trust misalignment that AI systems detect during cross-validation.
External Validation Depth
The External Validation score measures the depth of your sameAs chain. Direct sameAs links to Wikipedia, Wikidata, LinkedIn, Companies House, or official government directories produce the highest validation scores. Inferred links — where validation is implicit rather than explicit — produce lower scores. No sameAs links at all produces a validation score near zero. We also send HEAD requests to sameAs URLs during audits to verify they resolve — a sameAs link to a deleted Wikipedia page is a trust degradation signal.
Trust is Not Visibility
A highly visible but low-trust site is a different problem than a low-visibility, high-trust site. Visibility problems are solved by technical optimization: better schema, cleaner chunks, stronger entity definition. Trust problems are solved by consistency work: harmonizing entity data across pages, removing schema over-claims, building external validation signals. The Machine Trust Score is reported separately from the AI Visibility Score specifically because these are distinct failure modes requiring distinct remediation strategies.