Citation mining extracts and analyses real-time search grounding from LLM responses across Google, OpenAI, and Anthropic to track cited URLs, calculate brand share, and distinguish between selected vs. unselected sources.

sequenceDiagram
    autonumber
    actor User as User/System
    participant LLM as Grounded LLM
    participant Extractor as Citation Parser
    participant Resolver as URL Resolver
    participant DB as Postgres DB

    User->>LLM: Submit Prompt with Grounding Tool
    LLM-->>Extractor: Raw Grounded JSON
    Extractor->>Extractor: Extract URLs & Fan-Out Queries
    Extractor->>Extractor: Classify Selected vs Unselected
    alt URL is Vertex Search Redirect
        Extractor->>Resolver: Resolve Redirect (HEAD/GET)
        Resolver-->>Extractor: Clean Canonical URL
    end
    Extractor->>DB: Store Citations, Fan-Outs & Brand Mentions
    DB-->>User: Aggregate Share & Citation Profile