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.
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.
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.
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.
Why this page exists
Agentic readiness names a requirement that is becoming impossible to ignore: a website should no longer only be indexable, visible, citable, or machine-readable. It should be understandable as an action environment by systems capable of traversing an interface, selecting a target, filling a form, following a path, and executing an intention.
This hub is my personal entry point into the topic. It separates three regimes that are too often collapsed: visibility in search and answer systems, machine discoverability of documentary surfaces, and agentic readiness of interfaces. These regimes can reinforce one another, but they do not measure the same thing.
The founding distinction
A page can be highly visible in Google, correctly cited by an answer engine, properly listed in an llms.txt file, and still be weak for an agent that must act. Conversely, a highly robust agentic interface does not guarantee citation, ranking, or recommendation.
Agentic readiness begins when that distinction is accepted. The question is no longer only: “Can the system read this page?” The question becomes: “Can the system understand what it should do, what it should not do, which state it is in, which action is legitimate, and what consequence that action produces?”
Canonical entry point
Start with the definition of agentic readiness, then read it with these concepts:
- Agentic web: the website as an interpretable and actionable environment.
- Agentic navigability: the ability of an agent to traverse and manipulate an interface.
- Interpretable interface: coherence between visual intention, HTML structure, and machine exposure.
- Accessibility Tree: the programmable map of roles, names, states, and relationships.
- Machine readability: the corpus capacity to expose routes, canonicals, files, and reading conditions.
- Execution boundary: the separation between technical ability to act and legitimate authority to act.
Frameworks to read next
The agentic web readability framework states the technical controls: DOM stability, visual / DOM / Accessibility Tree coherence, actionable semantics, hydration risk, interaction determinism, visual hierarchy, and machine discoverability.
The AI visibility, machine discoverability, and agentic readiness matrix states the strategic distinction. It prevents llms.txt, Lighthouse, AI citations, structured data, sitemaps, SEO audits, and agentic navigability from being treated as if they belonged to a single score.
The position Search visibility is not agentic readiness fixes the doctrinal boundary: a search tool may surface or reward a source for reasons that prove nothing about its ability to be manipulated correctly by an agent.
Why Lighthouse changes the vocabulary
The integration of experimental Agentic Browsing audits into Lighthouse does not mean that llms.txt becomes a ranking factor in Google Search. It means that some machine readability and interaction signals are becoming observable inside standard developer tooling.
This is a category change. The topic is gradually leaving vague AI SEO opinion and entering verifiable controls: Accessibility Tree, visual stability, form exposure, discovery surfaces, coherence of intent, and a site’s ability to avoid forcing the agent to invent what the interface does not declare.
Editorial series
Read these articles in this order:
- Google, Lighthouse, and llms.txt: AI visibility or agentic readiness?
- Agentic readiness vs AI SEO: two regimes that should no longer be confused
- The Accessibility Tree as an interpretation map for the agentic web
- Why visual stability becomes a condition of machine action
- llms.txt: discovery, not governance
What this page does not promise
This page does not promise that an agent-ready site will rank better, be cited more, or be recommended more often. Those effects can be studied, but they should not be sold as automatic consequences. Agentic readiness is an architectural property, not a visibility promise.
The defensible promise is more sober and stronger: reduce action ambiguity, make intentions visible, stabilize paths, expose limits, and make the interface more readable for systems that may need to traverse it without immediate human supervision.
Reading rule
Never start with the tool. Start with the capacity: can the agent find, understand, distinguish, decide, act, abstain, and explain why? If the answer depends on fragile inference, agentic readiness is incomplete.