Relevance scores are calculated by submitting binary yes/no recommendation prompts across repeated samples and measuring the percentage of positive answers an LLM returns for a brand against specific entities or queries.

flowchart LR
    A[Brand + Query Target] --> B[Generate Binary Prompts]
    B --> C[Sample LLM N Times]
    C --> D[Parse Yes/No Responses]
    D --> E[Compute Relevance Score %]