The database maintains a set of normalized tables to track, verify, and curate brand statements across both recall modes and human evaluation:

Store Purpose
Claims registry Core claim record tracking origin, provenance flags, and truth status
Claim sources External web evidence proving or disproving a claim
Revision history Historical wording audit trail across curation runs
Ruling audit log Human verdict overrides (correct / incorrect / withdrawn)
Research findings Intermediate research facts linked to parent claims
Research citations URL-level citation metadata supporting research findings

The claims registry is the primary entity, tracking three boolean provenance flags—parametric recall, grounded research, and human review—alongside a verdict status of unverified, correct, incorrect, or gap. The claim-sources and ruling-audit stores hold supporting or contradictory web URLs and audit-logged manual human interventions, while the research-findings store links raw research into the centralized claim graph.