The free AI visibility audit at https://aideazz.xyz/api returned 100/100 for aideazz.xyz today. That score is built from 34 checks across four weighted groups. For a service business in Panama the same checks decide whether ChatGPT, Perplexity or Google AI Mode will recommend you or skip you.
What the audit actually measures
The audit runs 34 automated checks on one page. It returns a 0–100 score and an A+ to F grade. Each check reflects a signal that large language models use when they decide which service providers to surface for questions about Panama-based consulting, relocation help, or executive coaching.
The four groups and their weights are fixed: AI Crawler Access (25%), Structured Data (GEO) (25%), Answer-Readiness (AEO) (30%), Technical Foundation (20%). The groups are not marketing fluff; they mirror the crawl, parse, and ranking steps inside the models.
I run the same engine before any paid diagnostic. The $100 Quick AI Growth Operator Diagnostic at https://aideazz.xyz/pay/analisis-tecnico simply adds a written report plus a 45-minute video call that walks through the exact failures.
AI Crawler Access group (25% weight)
This group tests whether the major AI bots can reach and read the page at all. Eleven checks matter here.
Server returns a readable page (HTTP 200) is the first gate. robots.txt reachable follows. Then come the specific bot allowances: GPTBot for ChatGPT, OAI-SearchBot for ChatGPT Search, ClaudeBot for Claude, PerplexityBot for Perplexity, Google-Extended for Gemini, and CCBot for training corpora.
The audit also looks for llms.txt, XML sitemap available, and Page is indexable (no noindex meta/header). If any of these eleven fails, the other three groups become irrelevant for that model.
For a Panama service business this group decides whether your offer appears when someone asks “who helps executives move from Russia to Panama on an ordinary family budget.” No crawl permission equals zero visibility.
On aideazz.xyz every one of the eleven checks passed.
Structured Data (GEO) group (25% weight)
GEO here means Generative Engine Optimization, the structured signals that help AI systems map your business to real-world entities and answers.
Eight checks run. Structured data present (JSON-LD / microdata) is the base. Identity schema (Organization / Person / WebSite) tells the model who you are. Answer-rich schema (FAQ / Article / Product / Service) supplies the actual content the model can quote.
Entity links (sameAs to profiles / knowledge graph) connect you to Wikidata or LinkedIn so the model trusts the identity. Open Graph tags (title, description, image), Canonical URL declared, Language declared on , and Meta description (50–170 chars) complete the set.
A service business in Panama needs the language declaration and the sameAs links most. Without them the model cannot confidently tie your Spanish-language offer to the Panama GEO entity. The audit on aideazz.xyz passed all eight; the country-page commits from the last 48h added the exact question headings and sameAs markup the engine rewards.
Answer-Readiness (AEO) group (30% weight)
AEO stands for Answer Engine Optimization. This is the largest slice because it directly measures whether your page can become the quoted answer.
Nine checks. Title tag (15–70 chars) must exist and be concise. Exactly one H1 prevents model confusion. Section headings (H2/H3) structure content, and Question-style headings (FAQ signal) tell the model this page answers real queries.
Content depth (300+ words) is the minimum the audit accepts; I do not have deeper thresholds measured. Semantic HTML5 landmarks, Lists or tables (extractable facts), and Freshness signal (dates) round out the group.
For a service business the question-style headings and lists matter most. When Perplexity or Claude builds a “best relocation consultant in Panama” answer it pulls from pages that already look like FAQs. The aideazz.xyz audit passed all nine checks; recent commits added the exact headings that map to common Spanish-speaking country queries.
Technical Foundation group (20% weight)
This is the baseline. Seven checks that almost every modern site should pass yet many service businesses still fail.
Served over HTTPS, Response time under 1.5s, HTML under 1.5 MB, Mobile viewport meta, Image alt coverage ≥ 80%, Content present in raw HTML (not JS-only), No meta-refresh redirect.
The 1.5-second response and 1.5 MB HTML limits are hard cutoffs in the engine. A Panama service site that loads slowly or hides its text behind JavaScript simply does not exist for the AI crawlers. aideazz.xyz passed every check here; the PM2 processes that keep the stack live (nine online of nine) support the sub-1.5-second delivery.
How to read the overall score before you pay anyone
A 100/100 like the one aideazz.xyz received today means 34 of 34 checks passed and the weighted sum reaches the top bucket. That is rare. Most service businesses I audit sit between 45 and 70 even when they rank in traditional Google.
Read the score by group first. If AI Crawler Access is below 80 the page is invisible to at least one major model; fix that before touching anything else. If Structured Data (GEO) or Answer-Readiness (AEO) is the weak group, the model can see you but cannot confidently recommend you for Panama-specific service queries.
A single failing check can pull an entire group to zero because the engine treats the group as a chain. One missing sameAs link or one noindex header is enough.
The free audit at https://aideazz.xyz/api gives you the raw group scores and the list of failing checks in plain text. Use that output to decide whether the $100 diagnostic at https://aideazz.xyz/pay/analisis-tecnico is worth the time. The diagnostic only makes sense once you have seen the free result and still have more than two failing checks in the AEO or GEO groups.
What changes for a Panama service business
Panama-specific queries mix Spanish and English. The audit therefore weights Language declared on and Question-style headings heavily. Recent commits added one page per Spanish-speaking country plus the EspaLuz framing (“live your new country in Spanish”) that directly feeds the answer-rich schema.
Freshness signal (dates) also matters because service offers change with immigration rules. The audit only checks for any date; I do not have that measured at the daily level.
Entity links become critical. Without sameAs to a verifiable Panama or founder profile the model treats the business as ungrounded. The 12 commits in cto-aipa and 11 commits in aideazz over the last 48h all tightened those identity signals.
Common failure patterns I see
The most frequent zero in the AI Crawler Access group is missing llms.txt or a robots.txt that blocks one of the newer bots. In Structured Data the identity schema is often present but the sameAs array is empty. In Answer-Readiness the page has H2s but no question-style headings, so the model cannot match it to a user question.
Technical Foundation failures usually come from HTML over 1.5 MB or content hidden behind JavaScript. None of these are measured on aideazz.xyz today.
How the audit engine itself is kept current
The visibility-audit.ts engine sits at version 1.2.0. It is exercised daily against the live aideazz.xyz domain. The same PM2 process list that shows nine processes online (including n8n at 538 MB and cto-aipa at 202 MB) also keeps the audit endpoint responsive under 1.5 s.
Commits from the last 48h updated the wiki entry for AEO surfaces and refreshed the country-page markup. Those changes are visible in the 100/100 score but would not have been present two weeks ago.
The operator queue in NOW.md records that Cursor, Claude Code and the live processes share state only through HubSpot and the markdown files; the audit logic therefore reflects the exact production configuration.
Frequently Asked Questions
Q: Does a 100/100 score guarantee AI recommendations for Panama service queries?
A: No. The audit measures crawlability and answer formatting only. Actual recommendation also depends on the model’s training cut-off, the specific prompt, and competing pages. I do not have recommendation rate measured.
Q: How many of the 34 checks can fail before the score drops below 70?
A: I do not have that measured. The weighted groups make the exact threshold non-linear. A single failure in a 30% group hurts more than three failures in the 20% group.
Q: Why does the audit run on only one page?
A: The free endpoint tests a single URL so the business owner can paste their main service or country page. The paid diagnostic expands to additional pages when needed.
Q: Can I fix the llms.txt and sameAs issues without a developer?
A: Many of the checks are static files or meta tags. The free audit output lists exactly which files or tags are missing. Whether that is a solo task depends on your CMS; I do not have your stack measured.
Q: What is the connection between the 55193 restarts on algom-stream and the visibility score?
A: None directly. The audit measures the delivered page, not the internal restart counters. Those numbers appear in the same production environment but do not alter the 34 checks.
— Elena Revicheva · AIdeazz AI Lab · Portfolio · Professional Outlook (PDF)