Content Fetcher
One of AI Mode's two page-reading tools — it retrieves a batch of structured sources at once, versus browsing a single URL.
When Google’s AI Model needs to pull text from the web, it has two main ways of doing it. The first is browsing, which works like a human typing a single address into a web browser. The second, more systematic approach is the content fetcher.
Instead of opening one page at a time, the content fetcher takes a pre-defined list of multiple sources and retrieves them all at once in a single batch. In practice, the system runs a search first, gets back a list of result IDs, and then uses the content fetcher to grab all of those pages efficiently.
This is the crucial reading step that feeds the AI’s grounding process. It shows exactly how a generative system transitions from a broad list of search results to the specific page content it uses to formulate its answers. By working alongside other tools like document grounding and URL context, the content fetcher ensures the AI has direct, reliable access to the source material it needs.
Content fetcher is one of the two ways Google's AI Mode pulls text from pages. Where browsing takes a single URL — like typing an address into a browser — content fetcher takes a list of structured source references and retrieves them in batch, for a more systematic, workflow-driven kind of retrieval.
The distinction is in the inputs: browsing's browse(query, url) handles one arbitrary page, while content fetcher's fetch(query, source_references) takes many pre-identified sources (each with an id and type) at once. In practice, a system searches first, gets back several result IDs, then uses content fetcher to grab all of them efficiently.
For AI visibility it shows how a generative system moves from a spread of search results to the specific page content it will ground on. It's the reading step that feeds grounding, and it complements document grounding via the URL Context tool.