Parametric memory probing measures the unprompted associations and baseline facts an AI model preserves strictly within its training weights, often analyzed through token-level confidence and zero-search prompting. In contrast, grounded memory research systematically explores live web citations and page content via tools like Google Search and URL context across iterative rounds to build an evidence-backed factual profile. The platform’s Fact & Gap Checker automates the comparison between these two datasets, auditing every recalled claim against real-time research. Findings that validate training recall are marked as confirmed with supporting citation sources, factual inaccuracies are flagged as contradicted, and unprompted omissions are logged as brand gaps that highlight what the model fails to remember natively.