Scores and grades are Digital1010 house-rubric outputs from a point-in-time sample. They are not validated business outcomes, and a score or score movement does not establish traffic, citation, conversion, revenue, or lift. Unavailable measurements are excluded where the rubric says so; they are not converted into failures.
On 19 August 2026 these scanners were rebuilt after an independent adversarial audit of their own output. The audit found defects that changed scores, including sites credited with crawler access while refusing the crawler, and a sitemap read as 11 pages when it contained 1,933. The grade bands were unified across all three scanners on the same date.
Any score or grade issued before that date came from an earlier rubric and is not comparable to a score issued today. A re-scan may return a different number on an unchanged site. That difference is a change in how we measure, not a finding about the site. Every historical record was marked incomparable rather than quietly re-scored, and none of it is used in any published figure.
One homepage response plus scanner-requested site-root resources. Interior pages are not inspected. When the named crawler is unreadable but the house scanner is readable, crawler access is 0/20 and every other finding is explicitly limited to the scanner-served response.
Mozilla/5.0 (compatible; GPTBot/1.0; +https://openai.com/gptbot) D1010-ReadinessScan/1.0 (+https://digital1010.com/scan)The compound agent carries the GPTBot token because token-based bot protection otherwise does not reveal the named-crawler response, while also identifying Digital1010 and this scanner. It is used only for the named-crawler homepage probe. A server matching only the token can still interpret it as GPTBot; that limitation is not hidden. On a site owner's objection, Digital1010 will drop this probe for their domain on request.
The submitted entry page, its PageSpeed result and screenshot when available, up to 10 same-origin links discovered there, and one synthetic missing URL. Interior-page content is not inspected. Earned points are normalized over measured points when a listed source is unavailable.
A point-in-time sample of 12 configured prompts across four named models through OpenRouter with web search enabled. Provider errors are excluded. Model responses are non-deterministic, so a later score is another snapshot, not proof of improvement or decline.
Every AEO scan sends these twelve to each of the four models. {brand} is the name you submit; {service} comes from the category you pick. Two of the problem queries are written per vertical. Run them yourself and check our arithmetic.