We publish the whole rubric. A score you can't inspect is a number you can't act on — and every check below is something you can fix yourself, with or without us.
The AI Readiness Score answers one question: can AI engines read your website? It is a technical readability measure, scored out of 100 across 24 checks in six categories. It is deterministic — the same site scores the same today and tomorrow, with no language model involved in the scoring.
| 10 pts | JSON-LD structured data present Schema is the one place you state facts about yourself in a form models parse instead of infer. Without it, an engine is guessing from prose. |
| 5 pts | Entity schema (Organization / LocalBusiness / Person) Names the entity an AI answer would actually recommend. Without it you are a page, not a business. |
| 3 pts | WebSite schema Ties your pages together as one site so signals accumulate to a single entity. |
| 4 pts | Content schema (FAQPage / Article / BreadcrumbList) These are the blocks AI answers quote most directly. |
| 3 pts | sameAs profile links Corroboration. One source claiming who you are is a claim; five agreeing is a fact. |
| 4 pts | GPTBot allowed OpenAI's crawler — how ChatGPT learns your site exists. |
| 4 pts | ChatGPT-User allowed Live retrieval when a user asks ChatGPT about you right now. |
| 4 pts | ClaudeBot allowed Anthropic's crawler for Claude. |
| 4 pts | PerplexityBot allowed Perplexity's crawler — the most citation-driven engine. |
| 4 pts | Google-Extended allowed Gates Gemini and Google's AI Overviews. |
| 5 pts | Page loads (HTTP 200, under 3s) Crawlers work to tight fetch budgets. Slow or failing pages get abandoned, not retried. |
| 2 pts | Title tag The first thing every engine reads. |
| 3 pts | Meta description Your one-sentence answer to 'what is this site'. |
| 2 pts | H1 heading The page's claim, in one line. |
| 3 pts | Canonical URL Stops duplicate URLs splitting your signals across variants. |
| 5 pts | Readable text in the raw HTML (not JavaScript-only) Most AI crawlers do not execute JavaScript. A client-rendered site can look completely empty to them — this is the single most expensive silent failure. |
| 10 pts | llms.txt file exists The one file where you describe your site to language models in your own words. |
| 5 pts | llms.txt links to key pages A summary with no links tells a model what you are but not where to look. |
| 5 pts | FAQ / question-and-answer content AI answers are assembled from question-shaped content. If nothing on your site answers a question, nothing on your site can become an answer. |
| 3 pts | Open Graph tags Carries your name and image when you are cited or shared. |
| 2 pts | Structured lists and sections Lists and tables extract far more reliably than walls of prose. |
| 5 pts | sitemap.xml The index most crawlers start from. |
| 3 pts | robots.txt Without one, crawler behaviour is unpredictable. |
| 2 pts | Favicon Shows next to your name in AI answer citations. |
| 80+ | AI-ready Engines can read you. Focus shifts to whether they cite you. |
| 60+ | Nearly there A handful of fixes stand between you and full readability. |
| 40+ | Partially visible Some engines can read parts of your site. Most are guessing. |
| 0+ | Invisible to AI Fundamental blockers — usually robots.txt or JavaScript-only content. |
Four plain HTTP requests, about ten seconds: your homepage, /robots.txt, /llms.txt, and /sitemap.xml. We identify ourselves as SempiteTools/1.0, respect your server, and cache each domain's result for 24 hours.
We deliberately do not execute JavaScript — because most AI crawlers don't either. Scoring a site the way a real crawler sees it is the entire point; a headless browser would flatter client-rendered sites that models genuinely cannot read.
Two reasons, both self-interested and both honest. First, a scoring tool nobody can audit is marketing, not measurement — and we'd rather be quotable than mysterious. Second, everything on this list is genuinely fixable by a competent developer in an afternoon. If you fix it yourself, good: the web gets more readable and you were never going to be a client for a task that small. The work people hire us for is the part no rubric covers.
Rubric version 1.0 (July 2026). Checks and weights change as AI crawlers change; we'll note revisions here.
80 or above means engines can read you and the remaining work is about authority and citations rather than plumbing. Under 40 usually means something categorical is wrong — robots.txt blocking crawlers, or content that only exists after JavaScript runs.
No. Domain Authority estimates link popularity; Lighthouse measures performance and accessibility for browsers. This measures whether language-model crawlers can parse who you are and what you offer.
Because it's binary and absolute. A blocked crawler cannot read your site at any budget, and it is the most common self-inflicted cause of AI invisibility — usually added years ago by a plugin or a well-meaning developer.
Yes — the rubric is public precisely so it can be cited. An attribution link back to this page is appreciated.