Correction, 15 September 2026. Every per-check figure on this page was derived from a record of which checks scored below maximum, not of which features were absent. Those records hold no per-check points, so a claim like "57 of 58 had no content schema at all" was never something this data could support: a site that had content schema and fell short of the threshold was counted identically to one that had none. Separately, the 58 sites were scored under incompatible rubric versions, 49 of them under one our own records mark as not historically comparable. We withdraw the aggregate scores, failure rates and grades in full. The current verified cohort is 10 sites, below our publication floor of 20. Those are audit counts, not a finding about the market.
An AI readiness score can help you inspect a website. It cannot tell you whether an AI engine recommends the business. If your site supports inquiries and you want to check its technical foundations, start with the individual findings and the scope printed beside them.
What the scan reads
The scan requests the homepage both as our scanner and as a named AI crawler. When the crawler receives readable HTML, that response is used. When only the scanner receives readable HTML, the report labels the remaining findings as scanner scope and gives crawler access zero points. If neither request can read the page, it returns a failure outcome rather than manufacturing a score.
It also requests robots.txt, llms.txt, llms-full.txt and sitemap.xml, following supported sitemap indexes within scan limits. It does not search interior pages for their content or schema. Markup on your services page may therefore be outside the homepage check.
Some checks are conditional. For example, the robots.txt scoring check only applies when the named crawler can read the homepage. A check omitted from a report must not be counted as a pass.
Partial credit is different from absence
Each check has a maximum. Our stored failure list includes every check below that maximum, including partial credit. That list was the wrong basis for saying a feature was absent.
Content schema illustrates the distinction. Under the rubric dated 19 August 2026, the check awards three points per selected schema type found on the inspected homepage, capped at eight. Three or six points means matching markup was found in parsed data. It does not establish that every schema block is valid or that the markup accurately describes the page.
Read the finding before changing anything. Add only markup that is true of the page. A service page does not need unrelated article or product markup to chase our house score.
Useful checks you can take away
- Confirm the crawlers you want can receive real homepage content, and review the permissions in robots.txt.
- Keep the business name consistent in page text and identity markup. Link the profiles you control through sameAs; the scan checks for its presence, not your ownership of those profiles.
- Use genuine published or updated dates where appropriate. Keep sitemap dates tied to real page changes, rather than refreshing them to look current.
Some of these need a developer. The report is a starting list for that conversation, not evidence that every item applies to every page.
llms.txt and llms-full.txt are optional, experimental files. They earn points in this rubric; those points do not establish that an engine uses the files or will cite the site.
Measure citations separately
Our house score is not a validated measure of citations, traffic, conversions or revenue. It describes the checks we chose and the responses we inspected.
In our own Ahrefs citation-audit snapshot, re-verified on 27 August 2026, Perplexity showed one citation and the other six engines showed none. That is a dated observation, not today's count. The table and method are in our citation audit. That measurement answers a different question from a readiness scan.
If website inquiries matter to your business, use the scan to identify technical questions worth checking. Start with access, then identity and content accuracy. Read each finding before deciding whether it needs work.
Revised 14 September 2026. Cross-site comparisons remain withdrawn. No client or scanned domain is identified.