Relevance probing interrogates active LLMs using a strict yes-or-no prompt template that asks whether an AI platform would recommend a specific brand for a given topic. By repeating this inquiry across multiple sample passes (often up to 100 trials per model), the engine records raw responses in a dedicated probe-results store and calculates a percentage-based Query Relevance Score (QRS) representing the frequency of positive endorsements. The system tests these candidate pairings against curated queries or entities, excluding auto-generated fan-out tags, and aggregates results to identify specific topics where a brand is underperforming or deemed irrelevant.