agenciageo.lat

Why ChatGPT does not recommend your company, and how to find out

“¿Por qué ChatGPT no recomienda mi empresa?” The question usually arrives late, attached to evidence: a prospect mentioned a competitor that ChatGPT named. By then the loss already happened, quietly. The answer to a category question names three to five brands, and the brands left out do not even find out they were left out. No report shows it. No alert fires.

Here are the four reasons it happens, and the honest way to know which one is yours.

1. Retrieval: the model can only recommend what the open web says

ChatGPT does not know your company the way your sales deck knows it. For most questions it retrieves from sources it trusts, and the answer is assembled from them. If the trusted sources in your category never mention you, the model has nothing to recommend with. This is the most common cause and the most fixable, but only after you know which sources your category actually runs on.

2. The interchangeability trap: your competitor list is not the model’s

This one stings. Models routinely treat a premium brand as interchangeable with the supermarket private label next to it on the shelf. If your competitor list only contains the brands your sales team respects, every number you look at will be wrong in a way nobody notices. The model’s answer set is who you are actually competing against, and it is often not who you thought.

3. Sources you do not control carry more than your site

For a consumer brand, the sources that move the answer are typically retailer pages, review platforms, comparison guides and regulators. Most of them sit outside your direct control. That is not a reason to ignore them; it is the reason to know them. “We published a great page” and “the answer changed” are different events.

4. Variability: one ask proves nothing

Ask the same question three times and you can get three lists. Ask once, get a good answer, and you will believe you are fine. Ask once, get a bad answer, and you will panic. Both reactions are anecdotes. “Si hubieras tomado captura de la primera, habrías concluido que tu marca es invisible.” The same is true in reverse, which is why nobody should be making budget decisions off screenshots.

The honest way to know

Not a hunch, not one prompt, not a consultant’s opinion. A study:

  1. Build a prompt bank of real buyer questions in your category, and the questions never name your brand, because a named brand gets discussed and the measurement answers nothing.
  2. Ask them across several AI engines, several times each, so the numbers are stable enough to report.
  3. Read every response and count the same things always: mentions, recommendations, best-in-class signals, purchase paths, cited sources.

Here is what that looks like, from our invented sample (a sunscreen brand sold in Mexican pharmacies, measured against its actual answer-set competitors Bloqsol and Dermalux):

Marca inventada, competidores inventados, cifras inventadas. La estructura es real; los números no son de nadie.

Four numbers, each with its population printed next to it. Now the question “why does ChatGPT not recommend my company” has a where-to-start: which rate is low, which sources are carrying the answers, and who is inside the answer set instead of you.

That first run is the study: /auditoria-visibilidad-ia/.

Next step: The AI visibility study