This document is a placeholder for the spec for "Accessibility Tree Introspection", a part of the AOM project. For background, please refer to the Accessibility Object Model Explainer.

Phase 4: Computed Accessibility Tree

The Computed Accessibility Tree API will allow authors to access the full computed accessibility tree - all computed properties for the accessibility node associated with each DOM node, plus the ability to walk the computed tree structure including virtual nodes.

This will make it possible to:

• write any programmatic test which asserts anything about the semantic properties of an element or a page.
• build a reliable browser-based assistive technology - for example, a browser extension which uses the accessibility tree to implement a screen reader, screen magnifier, or other assistive functionality; or an in-page tool.
• detect whether an accessibility property has been successfully applied (via ARIA or otherwise) to an element - for example, to detect whether a browser has implemented a particular version of ARIA.
• do any kind of console-based debugging/checking of accessibility tree issues. react to accessibility tree state, for example, detecting the exposed role of an element and modifying the accessible help text to suit.

One implication of implementing this phase is that all browsers must agree on exposing the *same* accessibility tree for the same given webpage, otherwise differences will invariably lead to authored pages that work in some browsers but not others. Care must be taken to define this phase in such a way that maintains a high degree of compatibility between browsers, without exposing any internal details and nuances.