URL Context Tool
Google's Gemini capability that fetches and reads the content of a specific supplied URL, enabling document grounding.
Google has introduced a powerful capability called the URL Context tool, which allows its Gemini model to fetch and read the content of specific web addresses you provide.
It works much like a web browser. When you give the model one or more URLs, it visits those pages, extracts the main text and titles, and feeds that information directly back to Gemini. This is a major shift for Google, whose models were previously limited to standard web searches.
With this tool, Gemini can now perform document grounding. This means it can answer questions based on a specific page you point to, rather than just searching the open web. By combining this deep-reading capability with Google’s traditional search tool, Gemini can both find pages and analyze them thoroughly.
It is worth noting that while the Gemini app and its API can read live pages using this tool, certain versions, like AI Mode, have not supported it.
For creators and website owners, this change highlights a growing reality. Having clean, easily readable main content on your web pages is more important than ever, because AI models are now visiting your site and reading your content directly.
The URL Context tool is Google's capability that lets Gemini fetch and read the content of a specific web address you supply. Internally it works like a browse function: given one or more URLs, it visits each page, extracts the main text and title, and hands that back to the model to answer from.
This was a genuinely new step for Google, whose models were previously limited to the search-grounding tool alone. It enables document grounding — answering from a page you point at rather than from an open web search — and combines with Gemini's search tool so the model can both find and deeply read pages.
A practical note we flagged at launch: the Gemini app and API URL grounding can read live pages, but AI Mode did not. For SEOs it reinforces that clean, extractable main content matters whenever a model reads your page directly.