The engine uses an iterative Bayesian optimization loop to rewrite page snippets and content against specific LLM rankers, evaluating testable hypotheses mapped to a structured Rank Factor Class taxonomy.

flowchart TD
    A[Initial Content / Snippet] --> B[Seeder: Generate Hypotheses]
    B --> C[Classifier: Map Rank Factors]
    C --> D[Acquisition: Sampling / UCB]
    D --> E[Ideator: Apply Content Edit]
    E --> F[Ranker: Score vs Competitors]
    F -->|Reward Signal| G[Bayesian Posterior Update]
    G -->|Iterate| D
    G -->|Converged| H[Final Optimized Content & Brief]