Document Grounding
Grounding an AI answer in the content of a specific web page you supply, rather than in results from an open web search.
When you ask an artificial intelligence a question, it usually searches the open web for an answer. But there is another approach called document grounding. This is when the AI grounds its answer directly in the specific web pages you provide. You give the model a URL, it fetches the main text and title, and it answers using only that information.
OpenAI has offered this capability for a while, and Google recently introduced its own version called the URL Context tool, which lets Gemini read specific pages directly. For creators and website owners, this shift raises the stakes for machine readability. If an AI is going to fetch and read your page directly to answer a user's question, your site's main content needs to be clean and easy for a machine to extract. This direct page fetching is becoming a key part of how AI models retrieve information, sitting right alongside traditional web searches and advanced tools that pull from multiple structured sources at once.
Document grounding is when an AI model grounds its answer in the content of specific pages you hand it, rather than in results from an open web search. You supply one or more URLs, the model fetches and reads each page's main text and title, and it answers from that.
Google delivers this through its URL Context tool, an internal browse capability that prints a page's content and title for Gemini to read. OpenAI's models had this ability earlier; for Google it was a notable step beyond search-only grounding.
For content owners this raises the bar on machine readability: if a model is going to read your page directly, the main content needs to be cleanly extractable. It sits alongside broader grounding and the newer content fetcher path for pulling multiple structured sources at once.