How is this different from SEO?

A search engine returns a list of pages and the user chooses. A language model returns one answer and usually nobody checks where it came from. So you optimise not for a position in a list, but for being the sentence the model says.

Yes, though differently from a search ranking. You put a set of buying questions to the models and check whether the company appears in the answer, in what context and beside whom. Repeatedly, because answers change.

How long does it take?

Measurement takes days. Changing what models say takes months, because it needs content worth quoting and sources models read. Anyone promising faster is measuring something else.

AI search visibility (GEO) - measurement and improvement

Why this is a different discipline

In a search engine you compete for a position in a list and win by holding attention. In a model's answer there is no list. There is one sentence and a few names. There is also no click to tell you that you were left out - the loss is invisible from your analytics.

Models quote what is unambiguous, specific and attributable: a page that answers one question, gives checkable detail, and does not hide its content behind a script.

What we do

Measurement

A set of questions a buyer actually asks, put to several models, repeatedly. The result is not one number but a picture: where you appear, in what context, beside whom, and what the model says about you when it has no good source.

Content that can be quoted

A page about one subject rather than about everything. Answers to specific questions, specific numbers and dates, structured data in the markup. It is the same work as technical documentation, except the reader is a model.

Technical reachability

Content in HTML rather than assembled by script after load. Stable URLs, a sensible heading structure, an llms.txt. Dull, cheap things whose absence can exclude an entire site.

Honestly, the limits

Nobody controls what a model says, including us. What can be influenced is whether a good source exists and how readable it is. The rest is patience and repeated measurement.

Your customer stopped searching. They asked a model.