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.
Evidence layer
Probative surfaces brought into scope by this page
This page does more than point to governance files. It is also anchored to surfaces that make observation, traceability, fidelity, and audit more reconstructible. Their order below makes the minimal evidence chain explicit.
- 01Canon and scopeDefinitions canon
Definitions canon
/canon.md
Opposable base for identity, scope, roles, and negations that must survive synthesis.
- Makes provable
- The reference corpus against which fidelity can be evaluated.
- Does not prove
- Neither that a system already consults it nor that an observed response stays faithful to it.
- Use when
- Before any observation, test, audit, or correction.
Position
Search visibility is not agentic readiness. A source can be indexed, ranked, displayed, cited, reused in a generative answer, or listed in a machine-readable file without being ready to be used correctly by an agent.
This distinction becomes central as developer tools begin to expose Agentic Browsing signals. The fact that an audit detects llms.txt, the Accessibility Tree, visual stability, or tool declarations should not be confused with a Search ranking promise.
Principle 1: Search measures a discovery and relevance regime
Google Search, answer engines, and retrieval systems seek to find, select, summarize, cite, or display sources. Even when search becomes generative, it remains structured by indexing, retrieval, ranking, fan-out, and presentation regimes.
Those regimes answer questions such as: which source is relevant? which page can support an answer? which information is useful for the query? which page should be displayed? They do not prove that an agent can act correctly inside the interface.
Principle 2: llms.txt is an orientation surface, not action authority
llms.txt can help expose important surfaces. It can orient a system toward the right documents, hubs, files, or policies. That utility is real, but it primarily belongs to machine discoverability and machine readability.
An llms.txt file does not make a button actionable. It does not fix a field without a label. It does not stabilize a shifting layout. It does not prove that an action is authorized. It does not force a search engine to cite the site. It does not replace coherence between the visible corpus, machine files, and evidence.
Principle 3: Agentic readiness measures operational use
Agentic readiness asks something else: can an agent understand the interface, preserve context, choose the right target, avoid ambiguous actions, detect a limit, and cross only authorized execution boundaries?
This requires coherence between visual rendering, DOM, Accessibility Tree, links, forms, states, error messages, confirmations, structured data, internal linking, and policies. The page is no longer only an information source. It becomes a possible decision surface.
Principle 4: Lighthouse signals, it does not replace doctrine
Lighthouse’s experimental Agentic Browsing audits matter because they make some signals observable. They should not become a fetish. They do not measure the whole of interpretive governance, do not replace source audits, do not guarantee answer fidelity, and do not prove action authority.
Lighthouse can indicate whether a technical signal is present, absent, valid, or fragile. It cannot by itself say whether an organization has correctly bounded its commitments, exclusions, evidence, non-response rules, or responsibilities.
Principle 5: The three layers must remain separate
The diagnosis should always separate:
- Visibility: does the site appear, get cited, or get recommended?
- Discoverability: are the right surfaces accessible, routed, and hierarchized?
- Agentic readiness: can the interface be understood and acted upon without fragile inference?
Confusing these layers produces weak promises: selling llms.txt as a ranking lever, selling Lighthouse as an AI score, selling accessibility as an agent trick, or selling AI visibility as proof that the site is governed.
Doctrinal consequence
A site ready for the agentic era should be readable by humans, readable by machines, and readable as an action environment. These three requirements overlap, but they are not substitutable.
The human surface builds understanding. The machine surface organizes discovery, evidence, and canonicals. The agentic surface exposes actions, states, limits, and consequences. Dual Web must therefore integrate a third constraint: not only declaring meaning, but making action interpretable.
Normative rule
Never use the presence of llms.txt, a Lighthouse audit, structured markup, an AI citation, or a good Search ranking as standalone proof of agentic readiness.
Those signals can trigger analysis. They cannot conclude the analysis.