"AI visibility score" gets used by a growing list of tools, and it doesn't mean the same thing from one to the next — worth being upfront about before explaining our own.
What is an AI visibility score, honestly?
There's no single standardized definition. Different tools that use some version of this phrase measure different underlying things: some track how often a brand name is actually mentioned in live AI-generated responses (a measurement of outcomes), others — like ours — score how structurally ready a page is to be read, understood, and cited by an AI crawler in the first place (a measurement of readiness). Both are legitimate and useful; they're just not interchangeable, and a "72" from one tool and a "72" from another aren't necessarily comparable numbers.
How our score works
Our score comes from ai-visibility's AI Readiness Engine, the same open-source scoring engine that powers this site's own free scanner. It's six weighted categories, combined into a single 0-100 number:
| Category | Weight |
|---|---|
| Crawlability | 20% |
| Structure | 20% |
| Entity signals | 20% |
| Citation readiness | 15% |
| Content | 15% |
| Authority | 10% |
No machine learning and no hidden factors — every weight is a published constant, and a hard AI-crawler block (robots.txt disallowing every known AI bot, or a noindex tag) zeroes the whole score regardless of how well everything else checks out, since a blocked page has zero AI visibility no matter how well-structured it is. Full methodology and every individual check: the AI Readiness Engine docs.
How this compares to Semrush, Ahrefs, and Frase
Several established SEO platforms have added their own AI-visibility or AI-mention-tracking features — generally oriented around monitoring whether and how often a brand shows up across AI chat tools' responses over time, an outcomes-based measurement. Ours is structurally different by design: a readiness score for a specific page, checked against concrete, fixable technical and content criteria, plus (separately) a brand-measurement engine that does query real AI engines directly with your own API keys for mention/recommendation tracking. Treat readiness scoring and mention monitoring as complementary, not competing — a good readiness score describes whether a page can be cited; mention monitoring tells you whether it actually is, over time, across specific brand queries.
Check yours for free
- Web, no signup: crawlpod.com/scan
- Command line:
npx ai-visibility audit https://yoursite.com
Both run the identical scoring engine — pick whichever fits how you work.
What a good score looks like — real data
We ran this exact scoring engine against 50 real, live homepages in five categories (46 successfully scanned; 4 couldn't be fetched) — the full study is here. Worth an honest caveat before citing these numbers: that study used the classic 7-dimension analyze() engine (the predecessor to the current 6-category audit() described above), not the exact category breakdown in this post — but the underlying findings carry over directly, since both systems share the same core logic (a crawlability gate, an answer-placement/citation-readiness signal as the strongest differentiator).
From that dataset:
- Average score: 55.9/100. Median: 59/100. Most real sites cluster in a wide, unremarkable middle — not dramatically broken, not dramatically optimized.
- Answer placement (the precursor to citation readiness) was the strongest predictor of overall score (r=0.62) — and also the dimension most sites scored worst on (33.2 average). It's simultaneously the most commonly missed thing and the thing that matters most when done right.
- News sites scored lowest (47.4 average) — driven almost entirely by near-universal AI-crawler blocking (every news site in the sample blocked at least one AI crawler). SaaS sites scored highest (63.6 average), mostly because modern SaaS stacks tend to already ship clean structure and schema by default, not because of brand size.
llms.txtadoption: 43.5% (20 of 46 sites) already served a real one — higher than "new, obscure file" assumptions would suggest. Sites with it scored 61.3 on average versus 51.7 without — a real gap, though a correlational one from a single snapshot, not a controlled test.
Improve each category
One concrete lever per category, cheapest fixes first:
- Crawlability — confirm robots.txt explicitly allows major AI crawlers rather than assuming a default is a pass; run a scan to check all of them at once.
- Structure — one H1, an unskipped H2/H3 hierarchy, real semantic HTML landmarks.
- Entity signals — add Organization schema, and Person schema for byline authors.
- Citation readiness — state the page's direct answer in the first sentences, not buried after framing copy.
- Content — write sections that make sense quoted in isolation, without needing the preceding paragraph for context.
- Authority — visible author attribution and real, findable contact information.
See exactly where your own site lands: run a free AI visibility scan at crawlpod.com/scan, or npx ai-visibility audit https://yoursite.com — no signup required either way.