Relevance scoring repeatedly polls an AI model with binary yes-or-no prompts to determine if it recommends a specific brand for a given query or entity, converting the results into a percentage recommendation score.