Website observations
Repeated scans
Domains are observed repeatedly so that changes in source material, crawl conditions and interpretation can be separated over time.
SemanticRisk maintains a growing longitudinal dataset of website observations, model interpretations, extracted claims, score histories and reviewed semantic relationships.
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 pages provide benchmark summaries, rankings and a 14-day company history. Longer history, complete evidence, private-domain monitoring and operational alerts belong in paid monitoring and professional plans.
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.
Benchmark overview Company ranking Relationship calibration
Coverage snapshot updated 30 July 2026. Counts will be versioned as the public panel and reviewed calibration set expand.