Amazon's recommendation engine uses machine learning to personalize every customer's homepage, search results, product detail pages, and emails with items they are statistically likely to want to buy, based on their browsing history, purchase history, and the behavior of similar shoppers.

Amazon has operated AI-driven recommendations since the early 2000s, using item-to-item collaborative filtering — one of the foundational techniques in modern recommendation systems. The system is widely cited as responsible for approximately 35% of Amazon's total revenue. In September 2024, Amazon announced generative AI enhancements to its personalization layer: instead of generic labels like "More like this," the engine now generates context-aware labels like "Gift boxes in time for Mother's Day," built using Amazon Bedrock. The system evaluates personalized product descriptions and uses a secondary LLM as an evaluator to audit and refine outputs.