Inspectable agentic readiness reference

The inspectable agentic readiness reference exists to avoid a common drift: reducing a complex issue to a maturity score or marketing promise.

Pagup uses this reference model to connect public signals such as Lighthouse Agentic Browsing to a broader verification of the real site.

Principle

A site is not ready for agents because it has an llms.txt file, a good Lighthouse result or a few structured tags.

It becomes ready when its critical journeys can be understood and operated reliably by non-human systems.

That reliability must be inspectable.

Observed axes

The reference model checks:

• machine entry points: sitemap, internal links, context files, policies, proof resources;
• offer clarity: distinct services, audiences, deliverables, limits, use cases;
• multimodal coherence: visual rendering, HTML, DOM and accessibility tree;
• action semantics: buttons, links, fields, states, errors, confirmations;
• stability: absence of critical movement or change after hydration;
• form quality: labels, errors, context, post-submission message;
• proof: cases, method, limits, authority signals;
• governance: rules that prevent regressions after updates.

Useful outputs

The reference model produces concrete decisions:

1. fix a component;
2. rename a CTA;
3. connect proof to a service;
4. expose a page in llms.txt;
5. restructure a journey;
6. document an error state;
7. stabilize a form;
8. create a missing FAQ or case.

The output is not only a grade. It is a remediation map.

Acknowledged limit

This reference model does not claim to predict the behavior of every agent on the market. It verifies the properties that make a site more robust: clarity, stability, semantics, accessibility, proof and governance.

That approach avoids unverifiable promises. It makes it possible to say precisely what is solid, what is missing and what should be corrected.