People ask about AI visibility on Google, not inside ChatGPT - by a factor of 19.4 against 2.7 for ordinary business language.

Key findings

• Category vocabulary such as “generative engine optimization” and “AI visibility” appears 19.4 times more often on Google's AI answers than in ChatGPT; everyday business phrases appear only 2.7 times more often.
• For the 11 questions that literally name this category, recorded ChatGPT volume was zero and Google volume was 14,320 a month.
• All 11 of those category questions were answered with a web search, which means a page can reach the answer.
• “Generative engine optimization” shows a 23.8× gap between the two surfaces; “more customers” shows 0.9× - the same corpus, opposite results.

How this was measured

When

Measured 31 August 2026.

Instrument

A commercial corpus of captured AI assistant questions and answers, queried once per phrase per surface, reading the total match count rather than a sample.

Surface

Two surfaces, queried identically: ChatGPT, and Google's AI answers. The comparison is the point, so the filter is held constant and only the surface changes.

Sample

14 phrases - 10 in everyday owner language, 4 in category vocabulary - each measured on both surfaces.

Pre-registration

The everyday phrases act as the control band. They were chosen before the result was known, precisely so that a category-specific gap could be told apart from a corpus-wide one.

Why the control band matters

A raw finding that our category is more visible on one surface would prove nothing on its own. Any corpus is bigger on one surface than another, and the whole result could be an artefact of how the data was collected.

So every phrase was run on both surfaces with the same filter, and 10 of the 14 phrases are ordinary business language that has nothing to do with us. If the gap were an artefact, those 10 would show it too.

They do not. Their median gap is 2.7×. The category's is 19.4×. 2 phrases in the control band are effectively even, and 1 is slightly larger on ChatGPT.

The full comparison

Counts are matching records in the corpus, not monthly searches. Median ratio: everyday phrases 2.7× (n=10), category vocabulary 19.4× (n=4).

The sharper version of the same result

Narrowing to the questions that literally contain the category's own words makes the gap starker still. 11 such questions appear in the corpus: “generative engine optimization”, “answer engine optimization”, “ai visibility tool”, “ai search optimization” and their variants.

Their combined volume on Google's AI answers is 14,320 a month. Their combined volume on ChatGPT is zero.

And all 11 were answered with a web search rather than from the model's own memory - which is the part that matters commercially, because an answer built from a web search is an answer a page can get into.

A word about the word “GEO”

Anyone building a brand on the abbreviation should know what the corpus shows. On ChatGPT the term is dominated by geography, by slang, and by a currency in a video game. Only a handful of occurrences are ours.

It is a poor term to be found by on that surface, whatever it means inside the industry.

What this means if you are buying

• Today the demand for this category sits on an answer surface that a page can be cited into. That is a better position than it sounds, and it is the reason this site publishes measurements rather than product pages.
• Wanting to win inside ChatGPT does not move buyers there. On the evidence, buyers searching for this category are currently reachable through Google's AI answers, and a plan aimed only at chatbots is aimed at a room that is nearly empty.
• For a local business owner, the practical reading is the opposite one: he is not searching for this category at all, in either place. He is reached by someone contacting him, not by a page he looks for.

Honest limits of this measurement

• One corpus and one point in time. It records captured questions, not every question ever asked, and the balance between surfaces can change quickly.
• These are match counts within that corpus, not monthly search volumes, and they are not directly comparable to keyword tool figures.
• 14 phrases. A different set of phrases would shift the medians, though the size of the gap makes a reversal unlikely.
• The 11 category questions are those where the category's words appear literally in the question. A looser definition would include more questions and a larger volume; we report the version we can reproduce exactly.

Questions about this research

More measurements from the same instrument

• An AI answer to a business owner cites 8.7 sources
• How many companies can an AI assistant name at all?
• All seven measurements →

We run this measurement for a living. If you want it run on your own category and your own city, the free scan is the same instrument on one question, and the report is the full set. Nothing here is behind a form.

Every category we measure has its own page, with the questions customers put to an assistant in that category and what a reading costs. See all industries we measure →

Published 31 August 2026. Every figure on this page is recomputed from raw observation files by a single script, so any of them can be traced back to the readings behind it. If you are named here and believe a reading is wrong, tell us: we will run the measurement again and publish what it returns, whichever way it goes.