DocumentChunker
Chrome's internal Blink algorithm that recursively walks a page's DOM layout tree to group text nodes into semantically cohesive passages of up to 200 words.
Inside Google Chrome's rendering engine, a quiet tool called DocumentChunker is changing how the browser understands web pages. Instead of treating a page as one giant, flat block of text, this algorithm digs into the document's structure to group related text together.
It works by scanning the page layout, combining neighboring text and sub-sections into meaningful, semantically cohesive passages. Each of these chunks is kept to a maximum of about two hundred words. By doing this work directly on your device where the page is actually rendered, the browser can make sense of content in real time.
DocumentChunker acts as a vital building block for other Chrome systems. It helps generate annotated page content and creates the precise, passage-level units that artificial intelligence models use later on. When an AI needs to cite a specific source on a page, it relies on these exact grounding chunks to point to the right information.
What it is
DocumentChunker is an internal algorithm in Google Chrome's rendering engine (Blink) that recursively traverses a page's DOM layout tree, grouping sibling and descendant text nodes into semantically cohesive passages of up to 200 words. Instead of treating a page as one flat blob of text, it breaks the content into meaningful chunks that reflect the document's own structure.
Chunking on the device like this is a building block for understanding pages where they are actually rendered. It sits close to other Chrome content systems such as DomDistiller and the Content Fetcher, it produces the passage-level units described in Content Node and Annotated Page Content (APC), and it yields the sort of passage a model later cites as a Grounding Chunk.