Anthropic Taps the Web with New AI Search API

Claude Learns to Google (Sort Of)

Anthropic is expanding the capabilities of its Claude AI models with a new “Search” API, designed to pull in real-time, sourced web content into responses. Unlike traditional large language models (LLMs) that rely on static training data, this new API retrieves live information from the web, hoping to offer users more accurate, current, and grounded answers. The idea is to combat hallucinations—those infamous AI-generated errors—by anchoring Claude’s responses in actual articles, resources, or reports found online. As part of the rollout, Anthropic is partnering with web search provider DuckDuckGo’s Search API, using its content-surfacing stack called “DuckDuckGo Search & Surf.”

A New Layer for AI Answers

Anthropic frames the Search API as a solution to a deep problem with current AI assistants: opaque sourcing. Now, developers can give users contextual responses that include citations, letting them verify and explore where information originates. The API integrates directly into Claude workflows, and apps can programmatically request updated web knowledge alongside Claude’s language capabilities. Early use cases include research tools, AI chatbots with real-time reference capabilities, and educational software that can point users to trustworthy online sources rather than relying solely on pre-trained models. It’s a competitive push against rivals like Google’s Gemini and OpenAI’s web-connected GPT tools.