The Content Optimizer draws samples from each claim's Beta(α, β) posterior distribution alongside rank-factor-class diversity weighting to choose which content edit hypothesis to test next, which balances exploration with exploitation across optimization rounds.

flowchart TD
    A[Available Claims] --> B[Stage 1: Sample Class by Freshness]
    B --> C[Stage 2: Draw Sample θ from Beta Posterior]
    C --> D[Select Claim with Highest θ]
    D --> E[Craft Targeted Snippet Edit]
    E --> F[Score Variant Against SERP]
    F --> G{Rank Improved?}
    G -- Yes --> H[Update: α = α + 1]
    G -- No --> I[Update: β = β + 1]
    H --> J[Derive Status & Iterate]
    I --> J
    J -. Next Round .-> B