Website observations
Repeated scans
Domains are observed repeatedly so that changes in source material, crawl conditions and interpretation can be separated over time.
SemanticRisk retains website observations, model interpretations, extracted claims and comparable changes. Explore the published self-test, measurement methodology and fictional report example below.
These published research counts are historical context, not live coverage totals. The former public benchmark and company-ranking pages have been retired.
Domains are observed repeatedly so that changes in source material, crawl conditions and interpretation can be separated over time.
Model outputs are retained as structured records tied to jobs, scans and extracted claims.
Claims are normalized into families and comparable records while preserving the source wording needed for review.
Selected cross-model claim pairs are reviewed as equivalent, possibly equivalent, subset, superset, contradictory or unrelated.
Public resources include the dated SemanticRisk self-test, measurement methodology, calibration research and an explicitly fictional Unified Report. Comprehensive Review, Monitor and Compare & Monitor provide customer-specific assessments and, where included, recurring measurement and change history.
Robots rules, bot protection, authentication and intermittent errors can reduce available evidence.
Client-rendered or personalized content may differ from what the evaluation pipeline can retrieve.
Interpretations may change between runs even when normalized source content is unchanged.
Sector and overall comparisons depend on the companies included in the active benchmark.
Dated real self-test Fictional sample report Relationship calibration
Public resource links reviewed 16 October 2026. Historical counts and measurements retain their original dates; they are not a statement of current customer outcomes.