AI visibility score: what it is and how to improve it
An AI visibility score answers one question: how easily can an AI answer engine crawl, understand and recommend your site? It’s not your Google rank and it’s not your Domain Rating — it estimates whether models like ChatGPT, Perplexity and Claude can actually read you and have a reason to cite you. Here’s what goes into okro’s score, and how to move each part.
The foundation: can the machine reach you?
Before anything else, an engine has to fetch and parse your site. okro checks the crawl fundamentals — the same signals that make up most of the readiness score:
- robots.txt reachable and not blocking crawlers.
- sitemap.xml published so your pages are discoverable.
- llms.txt — a short, machine-readable profile that tells AI crawlers what you do, who you serve, and where the important pages are. This is the single highest-weight signal in the readiness score, and almost nobody has one yet.
- Enough crawlable text — real content, not a JavaScript shell the crawler sees as empty.
Fix these first. They’re cheap, they’re binary (you have them or you don’t), and they gate everything downstream — an engine that can’t read you can’t cite you no matter how good your product is.
Machine-readable positioning
Next, the model has to understand you without guessing. That means stating category, buyer, problem and outcome in plain language, and backing it with structure:
- Typed JSON-LD — Organization, WebSite, Product/SoftwareApplication, and FAQPage. Untyped or generic JSON-LD doesn’t carry the same weight; the entity type is what makes it usable.
- An H1 and meta description that say what you are and who it’s for without requiring inference.
- FAQ and comparison content — the format models lift from most when answering buyer questions.
Then: are you actually cited?
Readiness is necessary but not sufficient. The other half of the picture is measured, not inferred: okro runs your tracked buyer prompts against Perplexity, ChatGPT and Claude and records whether your domain is cited, mentioned, or absent — over time. A high readiness score with zero citations tells you the problem is off-site (corroboration, reviews, comparisons); a low readiness score tells you to fix your own site first.
How to improve it, in order
- Ship robots.txt + sitemap.xml + an llms.txt, and make sure your content renders without JS.
- Add typed JSON-LD and make your homepage state category/buyer/outcome plainly.
- Publish FAQ and comparison pages for your category’s buyer questions.
- Then monitor citations and treat each uncited buyer prompt as a task — okro can open the fix as a reviewable GitHub issue.
(Disclosure: okro is our product; the score model above is how okro actually scores a site.)