In the Content Optimizer, each generated claim is initialized as a Beta distribution \(\text{Beta}(\alpha, \beta)\) parameterized by the seeder's prior estimate \(\text{prior} \times N_0 + 1\) and updated via Bernoulli trials where ranking improvements increment \(\alpha\) and neutral or negative outcomes increment \(\beta\) across sequential rounds.