Amazon Bedrock now includes a built-in Web Search tool, giving developers a native way to connect foundation models to current web information without wiring in a separate search provider.
AWS says the feature is generally available and is meant for questions where a model’s training data is not enough, such as recent earnings, regulatory changes or time-sensitive facts. The tool is server-side and can be used during model inference, including through the OpenAI Responses API interface supported by Bedrock.
The product combines a continually refreshed web index with a knowledge graph that helps anchor factual entities and relationships. For enterprise teams, the practical appeal is less about novelty than procurement and architecture: fewer external vendors, fewer APIs to orchestrate and fewer separate security reviews.
Grounding does not eliminate hallucinations or replace source checking. It gives applications a managed path to retrieve current context before producing an answer, which can make chatbots, coding assistants and internal tools more useful when they need facts that changed after a model was trained.