This symptom appears when a site shows good technical results but remains difficult for AI agents to understand or use.
The site may have acceptable performance, indexable pages, some structured data, an llms.txt file, a satisfying Lighthouse result and still expose operational fragility.
Typical signs
The most common signs are:
- CTAs use generic labels without context;
- forms are visible but poorly exposed in the accessibility tree;
- confirmations or errors are not connected to the relevant fields;
- service pages look too similar;
- proof is not connected to the offers it supports;
- components change after hydration;
- menus, modals or filters are hard to interpret;
- machine files do not reflect the full site architecture.
In that case, the issue is not only technical. It concerns the relationship between interface, content, proof and governance.
Commercial risk
An agent may find the site but fail at the useful moment: choosing the right service, understanding the next step, preparing a request, filling a form or recommending an offer.
The site then creates a misleading impression of maturity. It looks ready because it passes selected tests. It remains fragile because it does not hold up across a complete journey.
Intervention direction
The answer is not to rerun Lighthouse only. Critical journeys must be audited through the Agentic Readiness Audit, then connected to agentic readiness, accessibility, internal linking, proof and machine surfaces.
The right goal is not to obtain a more reassuring score. It is to make the site more reliable when a non-human system has to act.