An AI accuracy figure needs a test for the specific task being advertised. In our synthetic Nimbus example, a website promises 98% accuracy at detecting AI-written content. The supplied test record reports 53% on a mixed sample of human and AI text. ClaimLens marks that claim as contradicted under its configured rules and keeps the reason attached to it. The same demo also has a supported fraud-detection claim. That result concerns a different task; it cannot supply the missing support for the content detector. A strong result elsewhere in the product does not resolve this gap. For a marketing review, ask for the record behind each sentence, including what was measured. ClaimLens organizes that comparison using supplied records. People still need to validate those records and decide what wording they support. #AIGovernance #MarketingCompliance #AIClaims