Ask an assistant who does something in your town and you get a paragraph naming two or three businesses. There is no list, no page two, and no explanation of why those three. If you are not one of them, you never find out the conversation took place.

It is worth understanding what is actually going on, because the popular assumption is wrong in a way that leads people to waste effort.

It is not summarising the top ten

The obvious guess is that the assistant reads the first page of Google and rewrites it. If that were true, ranking well would be enough, and it plainly is not: businesses holding position one for a term routinely go unmentioned when an assistant is asked the same question.

What actually happens is closer to research. The assistant works out what you are really asking, often breaking it into several narrower questions. It gathers material relevant to each, from what it retrieves at the time and from what it already holds about the subject. Then it writes an answer from the material it considers most reliable, and names businesses it has enough confidence to name.

Which means three things decide whether you appear

  • Whether it can read you. If your site blocks the crawlers that feed these systems, you are not in the material being considered. Nothing else can compensate.
  • Whether your pages answer the narrow question. The assistant is matching sub-questions to clear answers. A page covering everything about your service loosely will lose to a page answering that one question directly, even a page on a smaller site.
  • Whether other sources corroborate you. This is the part businesses underestimate. Being mentioned on the industry sites, directories and forums an assistant already trusts is often how it learns you exist at all. A business only ever described by its own website is a business with one source.

Why it favours specific over comprehensive

Because a question broken into parts gets matched against the clearest available answer for each part. Five clear pages, each answering one thing, offer five chances to be the clearest answer. One long page covering all five loosely offers roughly none, because there is nothing crisp to extract.

This runs against fifteen years of habit, where the advice was to build a comprehensive page. That advice was right for a ranked list. It is less right when the output is an assembled answer.

The country problem

One thing that catches out businesses outside the largest markets: without a location signal, these systems frequently answer about somewhere else entirely. Different prices, different currency, different companies. Your customer gets a confident answer about a market you do not operate in, and no local business appears at all.

Being the clearly local answer, with your area named in plain text rather than implied, is a real advantage here.

You cannot see any of this without asking

The uncomfortable part is that none of it is visible from a ranking report. The only way to know whether an assistant names you is to ask it, repeatedly, and record what comes back, including who gets named instead and which sources the answer was built from. That is what our AI visibility tracking does, weekly, on questions you choose.