The system evaluates content edits against a structured taxonomy of rank factors across core dimensions like substance and proof, using a custodian pipeline to surface emerging patterns from unclassified claims for human approval.