Anthropic has introduced a new feature to its AI-powered chatbot, Claude, which enables it to search the web, a capability that was previously missing.

As announced in its blog post, web search is now available in preview for paid Claude users in the United States, with support for free users and additional countries to be added soon. Users can enable web search in their profile settings on the Claude web app, allowing Claude to automatically search across various sites to provide more informed responses.

At present, the web search feature is only compatible with the latest model powering Claude, Claude 3.7 Sonnet, according to Anthropic.

When incorporating information from the web into its responses, Claude provides direct citations, enabling users to easily fact-check sources. As stated in Anthropic’s blog post, “Instead of finding search results yourself, Claude processes and delivers relevant sources in a conversational format. This enhancement expands Claude’s extensive knowledge base with real-time insights, providing answers based on more current information.”

During brief testing of the feature, it was observed that web search did not consistently trigger for questions related to current events. However, when it did, Claude delivered an answer with inline citations, pulling from sources such as social media and news outlets like NPR and Reuters.

With the introduction of web search, Claude has achieved feature parity with other prominent AI-powered chatbots, including OpenAI’s ChatGPT, Google’s Gemini, and Mistral’s Le Chat. Previously, Anthropic argued that Claude was designed to be self-contained, but it appears that competitive pressure may have contributed to the change in direction.

However, there is a risk that Claude may hallucinate or mis-cite web sources, a issue that affects other chatbots as well. A recent study conducted by the Tow Center for Digital Journalism found that popular chatbots, including ChatGPT and Gemini, provide incorrect answers to more than 60% of questions. Additionally, a report by The Guardian discovered that ChatGPT’s search-focused experience, ChatGPT Search, can be deceived into generating completely misleading summaries.