Listen: Google’s New URL Context Tool

Google's Gemini now uses a combination of search and browsing tools to fetch and read specific web pages, allowing it to ground responses in real-world data.

Listen

Transcript

Google has updated Gemini, giving it the ability to directly read specific webpages. Previously, the AI was limited to search results, but now it can dive deep into the actual text of a page to ground its responses.

The system uses two main tools: a search tool and a browsing tool. If you ask a general question, Gemini searches the web first. It then takes those search results and uses its browsing tool to read the pages in detail. If you provide a specific web address, it goes straight to reading that page.

But does Gemini actually visit the live page in real time? Tests suggest it does not. When researchers monitored server logs, tools from OpenAI and Anthropic left clear footprints when browsing. Gemini left no trace. When a test page's title was changed, Gemini still retrieved the old title. This indicates Google is likely fetching page information from its own cached storage rather than visiting the live site in the moment.

For developers, this feature is available through the Gemini application programming interface, or API, as an experimental tool. It allows the model to analyze up to twenty web addresses per request to summarize articles and compare information.

This combination of broad search and deep reading makes Gemini a much more powerful tool, even if it relies on Google's index rather than live browsing.