What Is an AI Visibility Score? (And What a Good One Looks Like)
An AI visibility score is a single number summarising how ready your website is to be crawled, understood and cited by AI answer engines. Ours runs 0–100 across five weighted categories. Here’s exactly how it’s calculated, what the grades mean, and the fastest route from any band to the next.
The five categories
Section titled “The five categories”| Category | Weight | What it measures |
|---|---|---|
| AI crawler access | 40 | Whether GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended and Bingbot can read your site |
| Structured data | 25 | JSON-LD presence, entity markup (Organization/WebSite), answer-friendly markup (FAQPage/HowTo) |
| Content signals | 15 | Descriptive title, substantive meta description, H1 |
| AI content guidance | 10 | An llms.txt file |
| Discoverability | 10 | XML sitemap |
Crawler access carries 40 points because it’s binary in effect: a blocked engine cites you exactly never, regardless of how good everything else is. The methodology page documents every check and fetch.
Reading your grade
Section titled “Reading your grade”- A (85–100) — Nothing structural is holding you back; citation share is now decided by content quality and entity trust.
- B (70–84) — Crawlable, but you’re making engines work: usually missing FAQ markup, llms.txt or entity schema. Each is a quick win.
- C (55–69) — At least one meaningful gap — often a blocked crawler or absent structured data. You’re being out-cited by readier competitors.
- D (40–54) — Multiple structural failures; AI engines can barely parse the site.
- F (below 40) — Effectively invisible to answer engines. Typically a blanket bot block plus a JavaScript-only frontend.
What the score can’t measure
Section titled “What the score can’t measure”A perfect 100 makes you citable, not cited. The score deliberately measures the controllable, technical layer. On top of it sit content quality (answer-first passages), topical depth, and third-party trust signals — which no crawler-side test can score honestly. Treat the score as the foundation gauge: fix everything it flags, then compete on content.
The standard raising sequence
Section titled “The standard raising sequence”- Unblock crawlers (robots.txt guide) — biggest single jump available, up to 40 points.
- Ship Organization + WebSite JSON-LD sitewide, then FAQPage on Q&A content (FAQ schema guide) — up to 25.
- Fix titles, descriptions and H1s on templates — 15.
- Publish llms.txt and a sitemap — the final 20.
Then re-run the free check — 5 checks a day means you can verify each fix as you ship it, and a monthly re-check catches regressions after deploys.