Watch: Annotated Page Content (APC)

Annotated Page Content, or APC, is a structured way to represent a webpage's layout and content. It organizes a page into a tree of nodes, allowing downstream systems to understand exactly what is on a page and how to interact with it.

Unlike the standard document object model, APC is built from the browser's layout tree. This means it only captures content that is actually rendered on the screen, keeping the data clean and efficient. Each node in the tree contains vital details, from visual coordinates and text styles to links, forms, and buttons.

This structured data can be converted into different formats, like structured Markdown or small text passages. It allows systems to identify and interact with page elements safely and reliably, even if the webpage changes dynamically.

Because webpages can contain sensitive data, APC is built with strict privacy and security guards. It strips out hidden passwords, tracks the origin of cross-site content, and respects paywall flags. Crucially, because it can contain private user information, APC data is designed to be temporary, and should never be stored beyond the immediate task without explicit consent.