Two acronyms search this page. Defined once, then straight to what matters.
- GEO (Generative Engine Optimization): the practice of making a brand more likely to appear in the answers of generative engines.
- AEO (Answer Engine Optimization): the same goal aimed at answer engines, the search results that respond directly instead of listing links.
What matters is whether AI recommends you. We measure that.
Whether you searched for an agency, a consultant or the acronyms, the question underneath is the same: when someone asks an assistant about our category, who gets named?
The debate, settled in one line
Some agencies sell GEO as a secret new framework. Others say it is hype and “just SEO”. Both are half right, and the halves do not cancel out.
Google’s own guide settles the optimization half: “optimizing for generative AI search is still SEO” (Search Central, “Optimizing for generative AI search”, updated July 2026: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide). The same guide names llms.txt, chunking and special schema as myths. So the skeptics are right: the optimization is mostly solid SEO, and anyone selling magic files is selling what Google says does not exist.
The skeptics stop one step short. Ask them how they know whether ChatGPT recommends you in your category this month, against which competitors, citing which sources. They cannot say. Measurement is the missing half, and it is not SEO. It is a study with a frozen prompt bank, repetitions and a codebook.
Method and proof
The method, in three steps: we build your prompt bank from your category and documented demand, we measure across assistants, with repetitions — the exact mix of assistant types is set per study — and we read every response and count the same things every time. Full method: /metodologia/.
What a report reads like, from the invented sample (a sunscreen brand sold in Mexican pharmacies, measured against its category competitors Bloqsol and Dermalux):
- Mentioned: 26%, 103 of 396 responses
- Recommended: 11%, 44 of 396 responses
- Named as the best: 3%, 12 of 396 responses
- With a purchase path: 7%, 28 of 396 responses
Every rate prints its denominator. A percentage with no population behind it is a decoration.
Sample one-pagers: the consumer example above and a B2B example, same structure.
The demo: we measure our own category
This page is the home of a self-referential study: who the assistants recommend for “agencia GEO México”, measured with the same frozen bank, the same repetitions and the same codebook as any client study, published with raw output included.
If the numbers look bad, they get published anyway. An evidence-first agency shows its own baseline.
Until the first run is published, the method is open for inspection: /metodologia/.
Built for Mexico
Almost everyone in this market quotes US studies. We generate Mexican data.
- Prompts in Spanish, written the way Mexican buyers actually phrase them, not translated US prompt lists.
- Local competitor sets, including the private-label trap: models often treat a premium brand as interchangeable with the supermarket brand next to it on the shelf. If your list only carries the brands your sales team respects, your share number is wrong in a way nobody notices.
- Country-level answer behavior. Assistants answer differently by market. A study run on US data does not describe what a buyer in Mexico City sees.
Spanish prompts, local competitors, local sources. That is what “agencia GEO México” should mean, and what the published demo will measure.
Your category is already being answered
Someone in your market asked an assistant the category question today, and the answer named somebody. The study tells you whether that somebody is you — measured on Mexican data, with every response saved behind the count.
Request the study. It opens with a conversation about the decision behind the question, and the scope follows from it.