Technical guide · original methodology
How to audit an AI-readiness claim
A repeatable review of scope, evidence, freshness, uncertainty, and optional capabilities.
Replace the badge with a question
AI-ready for which task, which observer, and which date? A site might expose a useful public article while its authenticated API is unavailable. A strong assessment starts with a bounded question and names the representation being tested. A broad readiness label without that context is difficult to falsify and therefore difficult to use.
Request the supporting record
For each claim, ask for the tested URL, timestamp, observer, method, result, and limitation through an appropriate access-controlled review. Configuration screenshots can support a configured-state claim. A successful request can support a response claim. Neither proves a ranking increase or an account conversion. Require the evidence to match the assertion; do not mistake access for publication permission.
Handle optional capabilities correctly
An informational business is not automatically defective because it lacks MCP, A2A, a public API, or llms.txt. Those should be assessed against declared use cases. In this method, not applicable is distinct from absent, and absence is distinct from an unsuccessful attempt to observe.
Look for motivated scoring
A report that gives every site a low score until it buys a service creates an obvious conflict. We recommend publishing the assessment criteria and making the free public evidence useful on its own. Separate the observations from any paid monitoring or implementation offer. Do not make a missing optional file a manufactured emergency.
Set a retest condition
Every proposed remediation should end with a repeatable acceptance test. The follow-up should preserve the original finding and record the new state, not overwrite the past. Useful research can show what changed and what remains unproven; it does not need to claim a causal revenue lift without a suitable study.
Sources and scope
Google: AI features and your website · Google: search spam policies
Provider documents support the referenced technical facts. The interpretation, examples and proposed checks are AIWebSignals methodology, not provider endorsement or measured industry results.
Review the assessment method and its limits · Request a correction privately · Read the JSON version