The AI Readiness Score grades any page from 0 to 100.

The AI Readiness Score measures whether a page is mechanically able to be found, read and quoted by AI search systems. It runs deterministic checks in five groups: can AI crawlers reach the page, is the content in the server-rendered HTML, is the document structured with one h1 and a clean heading outline, does the structured data parse and match the visible copy, and does the writing put answers first with the entity named. It is a property of your page, not a poll of what assistants currently say about your brand.

Because either failing zeroes everything downstream. A page that blocks the AI search crawlers cannot be retrieved at all, and a page whose words exist only after JavaScript runs is invisible to GPTBot, ClaudeBot and PerplexityBot, none of which executes JavaScript, as repeated measurement of hundreds of millions of crawler fetches has shown. Perfect schema on an unreachable page scores nothing, so the rubric weights reachability and readability above everything else, and says so.

Yes, in full, on this page. Almost no grader publishes its rubric, which makes most scores unfalsifiable marketing. RankX AI publishes every check, its weight and its pass condition, because the score is only useful if you can see exactly what to fix and verify the change yourself. The same rubric is applied to rankxai.com, and this site is expected to score 100 on its own checker.

Different instruments measuring different things. HubSpot’s grader and most AI visibility scores sample what assistants currently say or how a site performs generally. The AI Readiness Score measures page mechanics: whether the machinery of AI retrieval can reach, parse and quote this specific page. A famous brand can score poorly here while being talked about everywhere, and an unknown site can score 100 and still need content worth citing. The score tells you your page is ready, not that you are winning.

Three things, honestly. It scans a single page, not your whole site. It does not run a headless browser, so it measures the raw HTML exactly as AI crawlers do rather than what a browser eventually renders, which is the correct simulation of how retrieval works but means a heavily client-rendered page scores low even if Google can render it. And it uses no language model, so it checks structure and signals rather than judging whether your writing is good. Full-site scanning is what the RankX AI product does.

Yes. When fewer than two checks can be assessed, for example because the page is unreachable from our scanner, the honest answer is no score rather than zero. A zero would say your page failed everything; null says we could not measure it. The result marks each unassessed check and the score is computed only over what was actually assessed.

The controlled evidence says no. In 2026 Ahrefs added schema to 1,885 pages against thousands of controls and measured no citation movement on any AI platform, and tests show answer engines read visible text rather than JSON-LD. Schema still earns its place in the rubric for what it does do: Google rich results, Bing’s documented use of markup in its LLM pipeline, and entity disambiguation. The rule the score enforces is that every fact in your markup must also be in your visible text.

Because retrieval measurably favours it. An analysis of 30 million ChatGPT citations found 44% of cited passages come from the first third of the page, and a controlled split test that rewrote headings into question form measured a 12% organic uplift. The checks are mechanical versions of that: is there a direct answer near the top, does each section open with the answer rather than build to it, and does each section name its subject instead of leaning on pronouns that die when the passage is quoted alone.

Fix in the rubric’s own order. First unblock the AI search crawlers in robots.txt and at your CDN, which our AI Crawler Access Checker diagnoses precisely. Second, get your real content into the server HTML rather than behind JavaScript. Third, repair the document skeleton: one h1, no skipped heading levels, a title and meta description. Only then polish schema, dates and answer-first phrasing. The result lists your failed checks in weight order, so the top of the list is always the biggest win.

That is the standing bar for this site: rankxai.com scores 100 on its own AI Readiness Score, and the score is re-checked as the site changes. A company selling AI search readiness that failed its own mechanical audit would deserve exactly the scepticism that would invite. The rubric this tool applies is the same one RankX AI builds its own pages against.

After any template or platform change, and monthly otherwise. Scores move when frameworks change rendering behaviour, when a CDN changes bot rules, and when content edits quietly break structure, all of which happen without anyone deciding to fail an audit. Repeat scores inside 24 hours return the cached result instantly, so checking after every deploy costs nothing.

Free, no account, one page at a time, with fair-use rate limits per visitor and per domain to keep the scanner responsive. Results are cached for 24 hours per URL, so a repeat check is instant. Scanning a whole site, tracking scores over time and turning failed checks into fix tasks is what the RankX AI product does, and the free score is deliberately the same rubric so the numbers agree.