Skip to content

Article

The Accessibility Tree as an interpretation map for the agentic web

The Accessibility Tree is not only an inclusion requirement. It becomes an action map for agents.

CollectionArticle
TypeArticle
Categoryere agentique
Published2026-05-12
Updated2026-05-12
Reading time3 min

Governance artifacts

Governance files brought into scope by this page

This page is anchored to published surfaces that declare identity, precedence, limits, and the corpus reading conditions. Their order below gives the recommended reading sequence.

  1. 01Canonical AI entrypoint
  2. 02Public AI manifest
  3. 03LLMs.txt
Entrypoint#01

Canonical AI entrypoint

/.well-known/ai-governance.json

Neutral entrypoint that declares the governance map, precedence chain, and the surfaces to read first.

Governs
Access order across surfaces and initial precedence.
Bounds
Free readings that bypass the canon or the published order.

Does not guarantee: This surface publishes a reading order; it does not force execution or obedience.

Entrypoint#02

Public AI manifest

/ai-manifest.json

Structured inventory of the surfaces, registries, and modules that extend the canonical entrypoint.

Governs
Access order across surfaces and initial precedence.
Bounds
Free readings that bypass the canon or the published order.

Does not guarantee: This surface publishes a reading order; it does not force execution or obedience.

Discovery and routing#03

LLMs.txt

/llms.txt

Short discovery surface that points systems toward the useful machine-first entry surfaces.

Governs
Discoverability, crawl orientation, and the mapping of published surfaces.
Bounds
Incomplete readings that ignore structure, routes, or the preferred markdown surface.

Does not guarantee: A good discovery surface improves access; it is not sufficient on its own to govern reconstruction.

The Accessibility Tree is not new. But its role changes when interfaces are no longer only read by humans, but also explored by agents capable of acting.

In the agentic web, the Accessibility Tree becomes an interpretation map. It does not describe everything. It exposes what matters for understanding the roles, names, states, and relationships of an interface.

Why this layer matters

An agent may combine several sources: visual rendering, HTML, DOM after hydration, visible text, structured data, and Accessibility Tree. Each gives a different angle.

Visual rendering shows layout. HTML shows structure. The Accessibility Tree shows what the interface makes usable in terms of role and name.

If these layers contradict each other, the agent must arbitrate. That arbitration can become fragile.

An action map, not only a reading map

A button without an accessible name is not only a problem for a screen reader. It is also a badly declared action. A field without a label is not only user friction. It is an incomplete instruction. A modal without a title or focus management is not only a UX defect. It is a poorly governed state change.

In an agentic path, these details become conditions of action.

The danger of false affordances

Modern design often creates elements that look like actions without exposing themselves as actions. A whole card may be clickable without clear indication. An icon may carry the action without text. An element may become interactive through JavaScript without an explicit role.

For a human, context may be enough. For an agent, the system must connect what it sees with what is actionable.

The interpretable interface requires that coherence: appearance, code, and accessibility should carry the same intention.

Do not instrumentalize accessibility

This must remain clear: accessibility serves humans first. It is not a machine optimization tactic. But good accessibility makes the interface more explicit, and that explicitness also helps systems that read the action structure.

That is precisely why the topic matters. Agents are not an excuse for superficial accessibility. They reinforce the need for real accessibility, because a poorly named interface becomes fragile for everyone.

What to audit

An agentic audit should inspect:

  • buttons without accessible names;
  • generic links or links without context;
  • fields without programmatic labels;
  • open, closed, selected, or disabled states;
  • form errors associated with fields;
  • modals, menus, and accordions;
  • logical focus order;
  • important elements hidden from the tree.

Conclusion

The Accessibility Tree becomes a strategic surface because it makes the action map of the interface visible.

The agentic web does not only ask that sites be beautiful or fast. It asks that they be named, structured, and coherent. This is where accessibility stops being peripheral. It becomes a condition of interpretability.