The Grounded Validation system conducts multi-round live web research on a brand by querying search and scraping live content across recursive rounds to compile an evidence base from verified web sources. Operating in parallel with native recall testing, the engine runs the Fact & Gap Checker to evaluate whether live sources confirm, contradict, or uncover gaps against parametric memory claims. Extracted findings are systematically compared against baseline training recall, categorizing statements into validated facts, active hallucinations, or omitted brand knowledge, while linking every finding to verbatim cited URLs. Concurrently, a Veracity engine analyzes specific grounding segments against the source pages to compute a source fidelity score that measures how accurately the model preserves, loses, or distorts source information in its generated responses.