Methodology

Keep the evidence boundary visible.

Selection and timing

A fixed, purposive roster of 100 hostnames was submitted sequentially to the existing public scanner between 18:18:16 and 18:26:43 UTC on October 2, 2026. Ten thematic groups contained ten hostnames each. This was not a random sample or a representative ranking of the web.

AIWebSignals and Machine Realms were affiliated controls. Stripe and its documentation host, and Wikimedia and Wikipedia, are intentionally related surfaces; 100 hostnames does not mean 100 unrelated companies.

Returned evidence

99 usable Evidence Report payloads; 55 numeric technical scores; 44 reports with the score withheld; one unusable timeout result for hilton.com. The 205 retained finding rows include informational notes and inspection limits, not 205 proven business defects.

What a score means

The 0–100 number is the bounded public technical-readiness score. Broader evidence coverage is a separate ratio. For example, 28 known checks out of 48 applicable checks rounds to 58%, leaving 20 unknown. The six counterparty axes are not a validated end-to-end aggregate score.

Limits and counterexamples

No target-site account login, private telemetry inspection, purchase or authenticated agent task was performed. A suspected challenge may be a classifier error; a byte limit can omit later signals. Neither is proof that the target is broken. All material interpretations must preserve the hostname, timestamp and collector limitation.

How this publication was made

AIWebSignals Research used AI-assisted synthesis of the saved scanner payloads and checked quoted values and finding identifiers against those snapshots. Where a provider-specific rule is discussed, the article links primary documentation. Screenshots are actual report captures, resized for delivery rather than invented product interfaces. This is not an independent audit of AIWebSignals and is not presented as human-authored expert testimony.

Corrections and learning

Use the contact route to provide the exact hostname, report ID, observed discrepancy and a reproducible public source. Do not submit passwords, customer records or private prompts. Corrections require review and a dated update; a reader comment never silently overwrites the original observation.

Measurement and commercial boundary

Campaign links identify the content/channel, not an individual. Optional browser interaction measurement requires consent and honors GPC/DNT. A share click is not a confirmed recipient; a pricing click is not a purchase. Paid activation and connection configuration are separate existing server records. Attribution is association, not proof of causal sales lift.