We ran our own audit first.

Most agencies tell you what they could do for your AI visibility. We're going to tell you what AI says about us right now. The number is small and we're publishing it anyway. Across 120 queries on the four AI providers that matter — ChatGPT, Claude, Gemini, and Perplexity — XALA Studio gets named in answers 3.33% of the time. Four mentions out of 120. We just ran the methodology we sell on ourselves, and this case study is what we found.

The baseline

We picked thirty questions per AI provider — the kind of questions a Tijuana clinic owner, a real-estate broker, or a cross-border buyer might actually type. "How do I get my clinic to show up in ChatGPT." "Which studios in Tijuana work on AI search." "What's GEO." "Who optimizes for AI citations in Baja California." Then we ran each question through ChatGPT, Claude, Gemini, and Perplexity and counted how many times XALA's name appeared in the answer.

Total: 4 mentions out of 120 queries. 3.33% Share of Voice.

Per provider:

• Gemini — 4 of 30 (13.3%). Identifies the entity correctly when the question narrows enough that we're effectively the only available answer.
• ChatGPT — 0 of 30. Zero citations. Doesn't seem to know we exist.
• Claude — 0 of 30. Zero citations.
• Perplexity — 0 of 30. Zero citations.

That's worse than the headline number looks. Gemini is the only engine that found us, and only on questions where we were essentially the only available answer. The other three have no entity record at all.

The diagnosis

Three causes, in order of impact:

1. Brand collision. "Xala" is also a clothing line in Latin America, a co-working space in Madrid, and a personal brand on TikTok. When the model has to disambiguate "Xala" from a query, it tends to pick the more-cited entity — which isn't us. We score below them on the same baseline metric the model uses to weigh sources: Organization schema count, sameAs links to verified profiles, mentions in third-party publications.

2. No entity schema. Our homepage doesn't include an Organization schema with sameAs links, founding date, geographic coordinates, or service categories. AI crawlers extract entity facts from structured data first, then from prose. We're giving them prose only — and asking them to figure out the rest.

3. Thin cornerstone content. We have a homepage, a pricing page, an about page, and five journal articles. The journal is good but each article covers a different topic. There's no single canonical page that says, in machine-readable terms, "XALA Studio is a GEO consultancy in Tijuana that does X, Y, Z, for the following industries, with these deliverables."

The 30-day fix-list

We're shipping these in the next thirty days. Public scoreboard updated as each ships:

• Organization schema with sameAs pointing to our LinkedIn, Crunchbase, and Instagram. Plus geo coordinates, foundingDate (2026), knowsLanguage (es, en), areaServed (Baja California + cross-border San Diego).
• llms.txt at /llms.txt with the canonical signpost: who we are, what we do, which pages matter most, which we don't want crawled. Proposed standard from Jeremy Howard, 2024. Cost is one hour. Option value is real.
• Three cornerstone pages — one for clinics, one for brokerages, one for cross-border services. Each in the answer-shaped format: question heading, direct answer in the first sentence, supporting evidence with outbound citations, scoped FAQ at the end.
• Outbound citations on every claim. Every statistic links to the original source. Every methodology reference links to the paper. This article is the first one we wrote that way; you've already seen it.
• Service category disambiguation — a single sentence on every page that resolves "XALA Studio" to "Generative Engine Optimization consultancy" — not the clothing brand, not the co-working space, not the TikTok account.

Why we're publishing this

Two reasons.

First, the methodology only works if we believe in it. The fastest way to prove that is to apply it to ourselves and publish the results — good and bad. We're at 3.33% today. If we're at 30% in ninety days, the case is made. If we're not, you should ask why, and we should answer.

Second, the Mexican market for GEO is small. There are very few public baselines for Spanish-language AI search visibility — the field is dominated by US case studies on US sites. By publishing ours, we contribute one data point. As we add more — from clients who let us share, from sector studies, from the bilingual gap — they'll compound into something useful for everyone in this corridor.

The 5 Foundations, applied to us

The XALA methodology grades every site on five things. Here's how we grade right now:

• Readable — ✗ — no llms.txt, partial schema.
• Identifiable — ✗ — brand collision unresolved.
• Answer-shaped — partial — journal articles yes, cornerstone pages no.
• Citation surfaces — partial — some outbound links in journal, not systematic.
• Permitted — ✓ — robots.txt allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended.

Three failures, one partial, one pass. The fix-list above turns all five into passes within thirty days.

The next reading

We'll re-run the same 120 queries on August 20, 2026 — exactly ninety days from today. We'll publish the new SoV reading, the per-provider breakdown, and the delta. This page will become a living scoreboard.

If the methodology works, the number should at least double. If it doesn't, the methodology needs revising. Either way, you'll see the data.

Most agencies talk about what they could do. We just told you what AI says about us. Ask us what it'll say about you.