Machine Trust and GTL: How Trust Requires Honest Data State
You cannot build a trustworthy machine trust score on data whose completeness you do not know. GTL is the foundation that makes machine trust analysis interpretable and actionable.
Machine Trust analysis produces complex, multi-factor scores: entity credibility consistency, schema trust alignment, external validation depth, contradiction absence, and trust degradation resistance. Each of these can succeed, partially succeed, or fail to compute. Without GTL, a failed sub-computation silently produces a wrong number. With GTL, every machine trust score carries an honest declaration of what was computed and what was not.
Why Machine Trust Is Especially Vulnerable to Data Integrity Failures
Machine Trust analysis depends on more external dependencies than any other score tier. External validation analysis makes HEAD requests to sameAs URLs — which can timeout or return 404. Contradiction detection uses the Claude API — which can be unavailable or rate-limited. Trust degradation analysis requires a previous audit — which may not exist. Without GTL, each of these dependency failures produces a silent default (often zero), and the Machine Trust Score becomes an aggregate of real computations and silent zeros.
GTL in the Machine Trust Agent
The Machine Trust Agent is designed with GTL from the ground up. Each sub-score has its own failure handling: if external validation probes timeout, the externalValidationScore is set to null with reason "probe_timeout" — not zero. If contradiction detection fails, contradictionAbsenceScore is null with reason "claude_api_unavailable". The overall Machine Trust Score is computed from the non-null sub-scores only, with the denominator adjusted to exclude nulls. The GTL state of the overall score reflects the completeness of the sub-score set.
●A Machine Trust Score of 74 with state: "partial" and reason: "external_validation_excluded" is more useful and more honest than a Machine Trust Score of 58 where the 58 includes a zero for external validation that never ran. The user knows what 74 means and what to do next. They do not know what 58 means without the GTL context.
Trust in the Score Itself
A platform that produces machine trust scores without honest data state declaration is itself exhibiting the trust failure it claims to detect: presenting data with unknown reliability as if it were authoritative. GTL in the Machine Trust Layer is not just an implementation pattern — it is the platform embodying the same trust standards it measures. Every score SiteNexis produces carries GTL state, because SiteNexis is a machine trust platform, and machine trust requires honest data.