The audit of 34 checks across four groups returned 0/100 for the test page I ran yesterday. That is the constraint. A plastic surgery clinic in Panama will stay invisible to US patients asking ChatGPT or Google AI until every check in the "AI Crawler Access", "Structured Data (GEO)", "Answer-Readiness (AEO)" and "Technical Foundation" groups passes.
What the Assistants Actually Read
ChatGPT, Perplexity, Google AI Mode and Claude pull answers from pages that their specific bots can reach and parse. The visibility-audit.ts engine 1.2.0 I run checks exactly those bots: GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, plus CCBot for training data. If any returns blocked or empty, the page is invisible to that assistant.
The same engine confirms llms.txt is present, XML sitemap is reachable, and the page carries no noindex tag. Yesterday aideazz.xyz scored 100/100 on the full set; the plastic surgery clinic page I audited scored zero because it failed the first group.
The 25% That Blocks Most Clinics: AI Crawler Access
Server must return HTTP 200 with readable HTML. robots.txt must allow the bots listed above. The page must declare itself indexable. These 11 checks carry 25% of the total score. A clinic site that serves the content only after JavaScript runs or behind a login will fail the "Content present in raw HTML" check inside the Technical Foundation group and therefore never surface in AI answers.
I do not have that measured for any specific Panama clinic outside the test page. The pattern is consistent: most clinic home pages pass basic Googlebot but fail the newer LLM bots.
The 25% That Turns a Page Into a Recommendation: Structured Data (GEO)
JSON-LD or microdata must be present. It must contain Organization or LocalBusiness identity, Service or MedicalProcedure markup for the procedures offered, sameAs links to knowledge-graph profiles, Open Graph tags, a canonical URL, declared language, and a meta description between 50 and 170 characters. These 8 checks weigh 25%.
Without the Service schema that explicitly lists "plastic surgery" and the Panama address, the assistants have no structured fact to cite when a US patient asks for a clinic in Panama. The aideazz.xyz test page passed every item in this group and therefore appears when the right query matches.
The 30% That Makes the Page Answer-Rich: AEO Checks
Title must be 15–70 characters. Exactly one H1. H2 and H3 headings must structure the content. At least one question-style heading. Minimum 300 words of content. Semantic HTML5 landmarks. Lists or tables that an LLM can extract as facts. A freshness signal such as a last-modified date. These 8 checks are weighted 30% because they determine whether the page supplies a direct, quotable answer.
A clinic page that only shows before-and-after galleries without structured text will fail content depth and extractable facts. The audit engine measures raw HTML, not rendered images. The 100/100 result on aideazz.xyz came after we added question-style H2s and tables that match the exact phrasing patients use.
The 20% That Keeps the Page Alive: Technical Foundation
HTTPS, response time under 1.5 s, HTML size under 1.5 MB, mobile viewport meta, at least 80% image alt coverage, no meta-refresh, and content in raw HTML. These 7 checks are 20% of the score but act as gatekeepers. Fail any and the higher-weighted groups never get evaluated.
The test page that scored 0 failed here first, then cascaded. The page that scored 100/100 met every technical check before the structured and answer-readiness layers were scored.
How to Measure It Monthly
Run the free audit at https://aideazz.xyz/api once per month on the clinic home page and on the two highest-traffic procedure pages. Record the four group scores and the total. Track which exact checks flip from red to green after each change.
Commit the audit output to the same repository where you store the site. After every deployment, re-run the 34 checks. The aideazz.xyz run yesterday took 4229 ms and returned A+; a clinic page that stays at 0 will never appear in the assistants.
When the score reaches 80 or above, schedule the $100 Quick AI Growth Operator Diagnostic at https://aideazz.xyz/pay/analisis-tecnico. The diagnostic delivers the written gap analysis plus a 45-minute video call that maps the remaining checks to the exact schema and headings a plastic surgery clinic needs.
Frequently Asked Questions
Q: How many of the 34 checks must pass before ChatGPT will name a Panama clinic?
A: All 11 in AI Crawler Access plus the 8 in Structured Data (GEO) are non-negotiable. The test page that scored 100/100 passed every one; the page that scored 0 failed the first group and never reached the others.
Q: Does adding before-and-after photos improve the audit score?
A: Only if every image carries alt text (the ≥ 80% check) and the surrounding text supplies extractable facts in lists or tables. The audit engine ignores rendered galleries; it reads raw HTML. The 100/100 result included full alt coverage.
Q: Can I measure response time and HTML size without external tools?
A: The visibility-audit.ts engine 1.2.0 measures them in one run. The checks are "Response time under 1.5s" and "HTML under 1.5 MB". I do not have monthly averages measured for clinic sites.
Q: Will the same schema that works for aideazz.xyz work for a medical clinic?
A: No. The identity and Service schema must reflect LocalBusiness + PlasticSurgeryProcedure. The aideazz.xyz page used Organization and FAQ; the clinic page must swap those for the medical equivalents or it will fail the "Answer-rich schema" check.
Q: How often do the LLM bots re-crawl after a perfect 100/100 audit?
A: I do not have that measured. The freshness signal (dates) inside the Answer-Readiness group is one of the 34 checks; update it monthly and re-audit to confirm the page still passes all 34.