# Accessibility Tree

> The semantic representation of a page that assistive tech — and browser-driving agents — read instead of pixels.

_The Agentic Web Lexicon · /glossary/accessibility-tree · [JSON](/api/glossary/accessibility-tree) · [all The Agentic Web Lexicon](/glossary)_

- **term:** Accessibility Tree
- **category:** knowledge-memory
- **short_def:** The semantic representation of a page that assistive tech — and browser-driving agents — read instead of pixels.
- **long_def:** Derived from semantic HTML and ARIA and specified by the W3C Core Accessibility API Mappings, it exposes roles, labels and structure. Agents that drive a browser act through this tree, which is why accessible markup doubles as an agent interface: one investment, two audiences.
- **see_also:** webmcp, agent-experience
- **etymology_origin:** A construct of browser/user-agent accessibility APIs, normatively specified by the W3C in the Core Accessibility API Mappings (Core-AAM) and WAI-ARIA, developed by the W3C ARIA Working Group.
- **related_to:** webmcp, agent-experience, agentic-loop
- **contrast_with:** Unlike the DOM, which describes a page's full markup and presentation, the accessibility tree exposes only its semantics — roles, labels and states — which is what a browser-driving agent (and a screen reader) actually consumes.
- **example:** A browser-driving agent like Operator acts through the accessibility tree's roles and labels (per the W3C Core Accessibility API Mappings) rather than parsing pixels.
- **source:** https://www.w3.org/TR/core-aam-1.1/
- **status:** active
- **why_it_matters:** The accessibility tree means accessible markup IS an agent interface — one investment in semantic HTML/ARIA serves both assistive tech and browser-driving agents.
- **sameAs:** https://developer.mozilla.org/en-US/docs/Glossary/Accessibility_tree
- **bridge_entity:** agent-readiness
- **last_verified:** 2026-06-15
- **md_twin:** /glossary/accessibility-tree.md