The Algorithm Behind Amazon Rufus That Actually Decides Which Products Get Recommended

Amazon's search algorithm has evolved from keyword matching to a two-layer AI system consisting of Cosmo and Rufus. Cosmo, the backend, builds knowledge graphs to understand products, their purpose, and target audiences, while Rufus, the frontend, uses these graphs to answer shopper questions and provide recommendations. This shift means keyword-stuffed listings are losing visibility as the system now prioritizes understanding a product's purpose and relevance to shopper needs. Engagement signals, such as time on page, video views, and Q&A interaction, significantly impact how the AI understands and recommends products, with high engagement leading to better ad performance. Optimizing for Cosmo involves building a clear, accurate node in Amazon's Knowledge Graph by focusing on semantic clarity, modular listing structures, complete backend attributes, and a content strategy that addresses shopper concerns.