AWS Bedrock adds native web search for AI grounding

AWS Bedrock now includes native Web Search, a server-side capability that grounds foundation models like Anthropic's Claude or Meta's Llama in real-time web data without third-party APIs. The tool combines Amazon's continuously refreshed web index (billions of documents) with a built-in knowledge graph that handles factual queries with high confidence, reducing the small inaccuracies that occur when models stitch together fragmented text. Semantic snippet extraction pulls only the relevant passages from pages, optimizing the model's context window while keeping latency low.

Grounding becomes a built-in feature of Bedrock, activated with a single parameter in the OpenAI Responses API. The entire search lifecycle—query formulation, retrieval from Amazon's index, and response generation—happens server-side, with structured citations (URLs and titles) included in the API response. This eliminates the need for vendor onboarding, API orchestration, or data residency reviews, and ensures zero data egress by default, meeting enterprise compliance requirements out of the box.