DEJAN evaluates brand visibility by distinguishing between what an AI model recalls unprompted from its internal training weights and what it retrieves from live web citations.

flowchart TD
    subgraph Input["AI Model Input"]
        Q["User Query"]
    end

    subgraph Processing["Model Processing"]
        PM["Parametric Memory<br/>(Static Weights)"]
        GM["Grounded Retrieval<br/>(Live Search & Sources)"]
    end

    subgraph Output["Output & Audit Method"]
        R1["Unprompted Recall<br/>(Treewalker)"]
        R2["Cited Answers<br/>(Grounding Inspector)"]
    end

    Q --> PM
    Q --> GM
    PM --> R1
    GM --> R2

Referenced by