This page is about one tool’s data: the AI Visibility Checker. For how these scores compare with the other two tools, the distributions and the percentile lookup, see SEO Score Benchmarks — the methodology and privacy notes are there too.
Which citability signals are missing
On the GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) side, what we measure is not ranking but citability: can a language model fetch the page, work out who wrote it, and extract an answer at passage level. The items below are the links in that chain; the percentage is the share of scans in which the check did not pass.
| Check | Did not pass | Weight |
|---|---|---|
| llms.txt present | 70% | Partial |
| Publish / update dates | 66% | Partial |
| Author / byline | 60% | Partial |
| Organization / WebSite schema | 53% | Hard fail |
| About page linked | 52% | Partial |
| JSON-LD structured data | 48% | Hard fail |
| Question-style headings | 48% | Partial |
| Social / sameAs profiles linked | 41% | Partial |
| In-depth content (≥800 words) | 37% | Partial |
| Content in raw HTML (server-rendered) | 33% | Hard fail |
| Cites external sources | 25% | Partial |
| Concise, quotable passages | 22% | Partial |
The weakest AI-readiness area
Brand & entity strength leads by a clear margin: it was the lowest-scoring area in 47% of scans. Second is trust & freshness at 27%, third ai crawler access at 19%. The first measures missing Organization and WebSite schema, absent sameAs profiles and an inconsistent brand name; the third is agents like GPTBot, ClaudeBot and PerplexityBot being blocked in robots.txt without anyone noticing. The ranking says something useful: most sites are not blocking AI crawlers. Access is not the problem — what the model finds when it arrives is no identifiable identity to attach the content to.
Plenty of schema — just not the schema that says who you are
One detail stands out: JSON-LD structured data is entirely absent in 48% of scans, while the Organization/WebSite schema that identifies the site is missing in 53%. The gap is small but it points one way: not every site that adds schema adds the schema describing itself. The same pattern shows in the byline — no author in 60% of scans and no publish date in 66%. To a language model those add up to one thing: something is speaking, but there is no way to say who.
Frequently asked questions
What score do most sites get on an AI visibility check?
The median is 67/100 and 35% of scans reach 80+, but the distribution piles up at the extremes. Look at the grade mix: 473 scans scored F while 141 earned an A+. The 10th percentile is 20 and the 90th is 94 — a gap that wide means what this tool measures is not gradual but binary: either the basic machine-readable signals exist, or none of them do.
Which AI-readiness area is weakest?
Brand & entity strength is the weakest area in 47% of scans — a clear margin over the runner-up, trust & freshness, at 27%. It measures missing Organization and WebSite schema, absent sameAs profiles and an inconsistent brand name: in short, a site’s ability to tell a language model who it is.
My site has JSON-LD — why is my AI visibility score still low?
Because using schema and using the schema that describes you are different things. JSON-LD is entirely absent in 48% of scans, while the Organization/WebSite schema that identifies the site is missing in 53% — so some of the sites that do add markup are adding product, article or FAQ schema and skipping the markup that states who they are. To a language model, that is talking without ever giving your name.
The sample, the methodology, the anonymization and the CC BY 4.0 licence are all set out in the methodology section on the hub.