Accessibility Tree
A structured semantic representation exposed by the browser engine that records the role, label, and state of every interactive element on a page. Originally designed for assistive technologies such as screen readers, it has become a primary perception channel for AI browser agents.
Core Facts
Browser automation Agents locate webpage elements and simulate human operations through DOM parsing, visual recognition, or Accessibility Tree.
Compared to the raw HTML DOM, the Accessibility tree filters out a large amount of purely visual style information and retains semantically interactive elements, reducing token consumption while improving Agent perception accuracy.
Timeline (last 90 days)
Tools such as opencli drastically prune the DOM, first converting the page into an accessibility tree or semantic Markdown, discarding styles and decorative nodes, and retaining only semantic elements, thereby keeping a single call under 400 tokens.
Compared to screenshots, the accessibility tree has higher information density, consumes fewer tokens, and is unaffected by visual rendering, but some dynamic content or custom components may not be fully mapped.
All Facts (4)
Browser automation Agents locate webpage elements and simulate human operations through DOM parsing, visual recognition, or Accessibility Tree.
75%VerifiedCompared to the raw HTML DOM, the Accessibility tree filters out a large amount of purely visual style information and retains semantically interactive elements, reducing token consumption while improving Agent perception accuracy.
65%UnverifiedTools such as opencli drastically prune the DOM, first converting the page into an accessibility tree or semantic Markdown, discarding styles and decorative nodes, and retaining only semantic elements, thereby keeping a single call under 400 tokens.
50%UnverifiedCompared to screenshots, the accessibility tree has higher information density, consumes fewer tokens, and is unaffected by visual rendering, but some dynamic content or custom components may not be fully mapped.
50%