Relevance probing prompts active LLMs with a binary yes-or-no prompt asking whether a specific brand is recommended for a designated query or entity, aggregating multiple samples per model to compute a percentage-based relevance score.

flowchart LR
    A[Target Entity / Query] --> C[AI Model Probe]
    B[Target Brand] --> C
    C -->|N Samples| D{Recommendation?}
    D -->|Yes| E[Count Yes]
    D -->|No| F[Count No]
    E --> G[Relevance Score %]