GEO / AI-Search Audit — Clínica Longeva CDMX

GEO — Generative Engine Optimization — measures how well your clinic shows up inside the answers that AI systems write, not just inside the blue links of classic search. When a prospective patient in Polanco opens ChatGPT, Google's AI Overview, Perplexity, Gemini or Bing Copilot and types something like "best longevity clinic in Mexico City" or "clínica de terapia IV con vitaminas en Roma Norte," an AI assistant assembles a short, confident answer and names a handful of clinics. The whole game of GEO is: does it name you, does it describe you accurately, and does it link to you?

Today, mostly it does not. A 35 is a low-C / high-D result. It is not a sign that the clinic is weak — your reputation score (58) and your real-world standing (≈140 Google reviews, 4.6★) tell a much happier story. It is a sign that a genuinely good clinic is nearly invisible to the machines that increasingly mediate the first impression. That gap is the opportunity. Most of what holds the score down is fixable with structured content and technical hygiene, not with years of brand-building.

Two themes run through everything below and deserve to be stated up front:

• The bilingual gap is a GEO problem, not just a marketing one. Your Spanish content is reasonable; your English content is thin. AI assistants answer in the language of the question. When an international visitor or an English-first CDMX expat asks an English question, the model has very little of your material to draw on, so it reaches for competitors, directories, or generic content. You are losing a high-value, low-competition audience at the exact moment they are deciding.
• AI systems cite structure, not vibes. A warm, beautiful Spanish homepage that reads well to a human can be almost mute to a model if it lacks clear question-and-answer framing, named authors with credentials, structured data, and crawlable, factual service pages. The clinic's quality is real; it is just not machine-legible yet.

Composite GEO: 35 / 100. The two anchors dragging the average down — citability (28) and schema (30) — are also two of the fastest to move, which is why the score is more encouraging than it first looks.

1. AI citability — 28 / 100

Citability is the single most important GEO lever and your weakest. It asks a narrow question: if a model is writing an answer, how easily can it lift a clean, accurate, self-contained sentence or paragraph from your pages?

What we observed on the current site:

• Marketing prose over answer prose. Service pages lean on evocative, aspirational language ("redescubre tu vitalidad," "medicina de vanguardia") rather than direct, extractable statements a model can quote. A model cannot confidently reuse "vanguardia"; it can reuse "Clínica Longeva CDMX offers physician-supervised NAD+ IV therapy at its Polanco location, with sessions lasting 90–120 minutes."
• No question-shaped headings. Patients ask questions ("¿Cuánto dura una sesión de sueroterapia?" / "Is NAD+ therapy safe?"). Models strongly favor content where the heading is the question and the first sentence is the answer. Your pages are organized by service name and brand theme, not by patient question.
• Buried specifics. The facts that make you citable — protocols, session length, what conditions you do and don't address, who supervises, pricing posture, location and hours — exist in the team's heads and in PDFs/images, not as crawlable text. Anything locked in an image or a downloadable brochure is effectively invisible to most AI crawlers.
• The English deficit compounds it. Even where Spanish pages are reasonably citable, the English equivalents are stubs or machine-feeling translations, so for English queries there is almost nothing to lift.

Why 28 and not lower: a few Spanish service descriptions are factual enough to be quotable, and your Google Business presence gives models some baseline facts (name, area, rating). That floor is real but thin.

The move: rewrite the top 8–10 service and FAQ pages in an "answer-first" pattern — question heading, a 40–60 word direct answer, then supporting detail — in both languages. This one change tends to lift citability faster than anything else.

2. Brand authority — 50 / 100

Brand authority is how well AI systems recognize Clínica Longeva CDMX as a distinct, real entity and how much corroborating evidence exists about it across the web. Models don't just read your site; they cross-check it against directories, the knowledge graph, press, professional listings and review platforms before deciding to name you.

Strengths:

• A solid, consistent Google Business Profile with ~140 reviews at 4.6★ is a genuine authority signal — it tells models you exist, you're active, and patients rate you well.
• The clinic name is distinctive enough to be disambiguated (it is unlikely to be confused with a national chain).

Gaps:

• Thin entity footprint. Beyond Google and the clinic's own site and Instagram, there is little independent corroboration — few authoritative directory listings, limited press, no clearly structured "about the clinic / about the physicians" entity that ties everything together. Models like triangulation; right now most signals point back to a single source (you), which models discount.
• No knowledge-panel-grade consistency. Name, address, phone, hours and category are not perfectly consistent across the handful of places you do appear. Inconsistent NAP data makes a model less sure it's the same clinic, which suppresses confident citation.
• Physician identity is under-leveraged. The clinic's credibility lives largely in its doctors, but the doctors are not established as web entities (consistent bios, credentials, professional profiles) that a model can attach to the clinic.

This is the dimension where reputation (58) and GEO meet: you have earned authority in the real world that has not been translated into machine-readable corroboration.

3. Content E-E-A-T — 38 / 100

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is the framework AI systems and search engines use to decide whether health content is safe to surface. For a medical clinic, this bar is higher than for almost any other business, because longevity / regenerative / IV content sits squarely in "Your Money or Your Life" territory, where models are deliberately conservative.

Where you stand:

• Experience & Expertise (implied, not shown). The clinicians clearly have it, but the site rarely demonstrates it: pages are mostly unsigned, there are no named author bylines with credentials (cédula profesional, especialidad, board memberships), and there is little "we have treated X patients / here is how we actually run a protocol" first-hand framing.
• Authoritativeness (weak external). Few citations to recognized sources, no links to or from authoritative health bodies, and limited evidence the clinic participates in the wider professional conversation.
• Trustworthiness (mixed). Reviews help, but the site is light on the trust scaffolding models look for: clear medical disclaimers, transparent descriptions of risks/contraindications, who supervises each therapy, last-updated dates, and an honest scope of what a given therapy can and cannot do. (This last point is also where E-E-A-T and COFEPRIS / PROFECO compliance overlap — see the issues section.)
• Bilingual asymmetry again. English pages carry almost none of the E-E-A-T scaffolding the Spanish pages partially have.

The move: put real, credentialed humans on the page. Named clinicians, photos, cédula/especialidad, a short "reviewed by Dr. ___ on [date]" line on every clinical page, plain-language risk/eligibility sections, and dated content. This simultaneously raises E-E-A-T and reduces compliance exposure.

4. Technical GEO — 45 / 100

Technical GEO asks whether AI crawlers can actually reach, render and parse your content. This is your strongest GEO sub-dimension — the foundations are mostly sound — but several quiet issues cap it at fair.

Observed:

• Reachable and reasonably fast. Core pages load acceptably on desktop and mobile; HTTPS is in place. Good.
• Rendering risk. Portions of the site appear to depend on client-side JavaScript to display key content (service details, some text). Several AI crawlers fetch raw HTML and do not execute JavaScript well or at all — so content that only appears after JS runs may be invisible to them. Anything important should be present in the server-rendered HTML.
• Crawl signals. robots.txt and sitemap hygiene is incomplete (see the AI-crawler section), and there is no llms.txt — the emerging convention for telling AI systems what your site is and where the good content lives.
• Structure & internal linking. Flat, thin internal linking means a crawler has a harder time understanding which pages are your "money" clinical pages versus brand/landing fluff. Clear hub-and-spoke structure (a longevity hub linking to each therapy page) helps models map your offering.
• Hreflang missing. With two languages, the lack of proper hreflang annotation means engines aren't told which page is the ES version and which is the EN version of the same content — directly worsening the bilingual problem.

5. Schema markup — 30 / 100

Schema (structured data, usually JSON-LD) is how you hand machines pre-digested facts. It is one of the highest-leverage, lowest-effort GEO fixes, and right now it is almost entirely absent.

What's missing:

• MedicalClinic / MedicalBusiness — the core entity type that tells engines what you are, with address, geo coordinates, opening hours, languages spoken (availableLanguage: ["es","en"] — note the bilingual signal), and contact points.
• Physician — structured profiles for each named clinician, tied to the clinic.
• MedicalProcedure / Service — one per therapy (IV/sueroterapia, NAD+, regenerative protocols, longevity assessments), with plain descriptions.
• FAQPage — your patient questions in structured form; this is frequently the exact content models lift into AI answers.
• Review / AggregateRating — to surface your 4.6★ standing in a machine-readable way.
• BreadcrumbList — to clarify site structure.

A handful of generic tags may exist, but nothing health- or clinic-specific. Adding the schema above is mostly a developer task measured in days, not months, and it tends to move both this sub-score and citability.

6. Platform optimization — 30 / 100

Each AI engine selects and cites local clinics a little differently. Generic optimization helps everywhere; platform-aware optimization wins the specific surfaces your patients use. You're currently tuned for none of them. The per-platform readiness table below breaks this down.

Before any platform can cite you, its crawler has to be allowed in and able to find the good stuff. Here's the current picture and the target state.

Findings:

• No llms.txt. This plain-text file (served at /llms.txt) gives AI systems a curated map: who you are in one paragraph, your key service pages, your FAQ, your contact and location, in both languages. Its absence is a missed, near-free signal. An outline is below.
• robots.txt is permissive-by-omission, not permissive-by-intent. Nothing is actively blocking AI crawlers today, which is good — but because the major AI agents aren't named, you have no control and no clarity. We recommend explicitly allowing the crawlers above so that future site changes (or a developer "tightening" robots.txt) don't accidentally lock AI systems out.
• Sitemap not referenced from robots.txt, and the sitemap itself appears partial — meaning crawlers may never discover some service pages.
• A practical caution: do not block the AI crawlers as a "privacy" reflex. For a clinic that wants to be discovered, blocking GPTBot/PerplexityBot/Google-Extended is the digital equivalent of taking your sign down. Allow them; control what you publish instead.

llms.txt — example outline (to be built, ES + EN)

# Clínica Longeva CDMX

> Physician-led longevity, regenerative and IV-wellness clinic in
> Mexico City (Polanco & Roma). Bilingual care (Español / English).
> Personalized, medically supervised protocols.

## About
- /about — Clinic overview, philosophy, facility (ES)
- /en/about — Clinic overview (EN)
- /equipo-medico — Physicians, credentials (cédula, especialidad)
- /en/medical-team — Physicians (EN)

## Services
- /servicios/sueroterapia-iv — IV therapy / vitamin drips
- /servicios/nad — NAD+ protocols
- /servicios/medicina-regenerativa — Regenerative protocols
- /servicios/evaluacion-longevidad — Longevity assessment
- /en/services/... — English equivalents of each

## Patient information
- /preguntas-frecuentes — FAQ (eligibility, duration, safety, pricing posture)
- /en/faq — FAQ (EN)
- /aviso-medico — Medical disclaimer & scope of care

## Contact
- /contacto — Address (Polanco / Roma), hours, phone, WhatsApp, booking
- Languages: Español, English

Keep it factual and current; the file is only useful if it stays in sync with the live site.

How ready is the clinic to be named and cited when a patient runs the queries below on each engine? Representative queries we modeled: "longevity clinic in Mexico City," "clínica de sueroterapia cerca de mí Polanco," "best IV therapy Roma Norte CDMX," "NAD+ therapy Mexico City English-speaking doctor," "medicina regenerativa CDMX."

Cross-platform pattern: the same three things move every engine — (1) answer-first, bilingual content; (2) clinic/physician/FAQ schema; (3) a stronger, consistent off-site entity footprint. You do not need a different strategy per platform so much as the shared foundation plus a few platform-specific listings (Bing Places, Google Business optimization, a couple of authoritative MX health directories).

The bilingual angle, specifically: for the English queries above, every platform currently tends to reach for English-language directories, generic "medical tourism" content, or competitors — because you've given it almost nothing English to cite. This is the rare case where modest effort (a clean English mirror of your top pages) opens a high-value audience that very few CDMX longevity clinics are competing for online.

Critical (suppressing citation right now)

• No medical/clinic schema. Engines lack machine-readable facts about what you are, who your physicians are, and what you offer. (Schema 30)
• Content is not answer-first / not citable. Marketing prose instead of quotable, self-contained answers. (Citability 28)
• English content is a stub. A high-value, low-competition audience is effectively unserved by AI on every platform. (Citability / E-E-A-T / Platform)
• Advertising-claim & disclaimer gaps with compliance overlap. Some service language leans toward outcome/benefit claims for longevity/regenerative/IV therapies without clear scope, eligibility, risk and supervision context. Under COFEPRIS (health-product & advertising rules, incl. permiso de publicidad for certain health claims) and PROFECO (misleading-advertising / consumer protection), this is both a compliance flag and an E-E-A-T drag — AI systems down-rank unqualified health claims. Fixing it helps both. (Compliance 33 / E-E-A-T 38) — marketing read, not legal advice.

High

• No llms.txt and AI crawlers not explicitly allowed — no control, no map. (Technical 45)
• JS-dependent rendering may hide key content from non-rendering crawlers. (Technical 45)
• No named, credentialed authors / physician entities. (E-E-A-T 38 / Brand 50)
• Missing hreflang — ES/EN versions not properly linked, worsening the bilingual gap. (Technical 45)

Medium

• Inconsistent NAP across the few listings you have. (Brand 50)
• Partial sitemap, not referenced in robots.txt. (Technical 45)
• Weak Bing Places / non-Google directory presence. (Platform 30 / Brand 50)
• No FAQ content in either language — the exact format models love to cite. (Citability 28)

Low

• No last-updated dates on clinical pages (a trust + freshness signal).
• Thin internal linking between the longevity hub and individual therapy pages.

These are the moves that tend to shift the GEO score fastest for the least work:

• Publish a bilingual llms.txt using the outline above. Hours of work; immediate clarity signal.
• Add core schema — MedicalClinic + Physician + FAQPage — to the homepage, team page and top service pages. Days of developer time; moves schema and citability.
• Rewrite your top 5 FAQs answer-first, in ES and EN — duration, safety/eligibility, what to expect, who supervises, pricing posture. Mark them up as FAQPage.
• Explicitly allow AI crawlers in robots.txt and reference a complete sitemap. Near-zero effort, removes risk.
• Stand up clean English mirrors of your 6 highest-value pages (home, about, team, top 3 services) and wire hreflang. The single biggest bilingual unlock.
• Make NAP identical everywhere and claim/optimize Bing Places. Quiet, cumulative authority.

Phase 1 — Foundations (weeks 1–4): the six quick wins above. Goal: make the clinic machine-legible and crawler-accessible in both languages. Expected effect concentrated in schema, technical and citability sub-scores.

Phase 2 — Citable depth (weeks 4–8): rewrite the top 8–10 service pages answer-first in ES + EN; add named, credentialed physician bios with cédula/especialidad and "reviewed by ___ on [date]" lines; add plain-language risk/eligibility/scope sections that double as COFEPRIS/PROFECO hygiene; build the longevity hub-and-spoke internal linking. Goal: lift citability and E-E-A-T, which feed every platform. (Compliance language = marketing read, not legal advice; have qualified MX counsel review claims and any permiso de publicidad needs before publishing.)

Phase 3 — Authority & corroboration (weeks 8–12): build the off-site entity footprint — authoritative MX health/clinic directories, consistent physician professional profiles, and a small amount of genuinely useful published content (e.g., evidence-aware explainers) that other sources can reference. Goal: give models the triangulation that turns "exists" into "trustworthy enough to name."

A necessary honesty note: we never promise rankings or specific AI placements — no one can, and anyone who does is selling certainty that doesn't exist in this space. AI engines change their selection logic frequently, and geo-tagging or schema is not a guaranteed ranking factor. What we can commit to is making the clinic dramatically more legible, accessible and trustworthy to these systems than it is today, and measuring the movement honestly month over month. Given how strong your real-world reputation already is, that legibility is the missing link — not the underlying quality.

GEO scoring synthesizes AI-citability heuristics, crawler-access and rendering checks, schema/structured-data inspection, E-E-A-T review, and per-platform (Google AI Overviews, ChatGPT, Perplexity, Gemini, Bing Copilot) readiness modeling against representative local + service queries; figures are illustrative composites for a fictional subject, benchmarked to industry norms, not guaranteed outcomes; compliance commentary (COFEPRIS / PROFECO / NOM) is a marketing read, not legal advice.

← Back to the full report