Search Engine Land reported on 30 July 2026 on a study by geoSurge finding that AI models search for brands they already know far more often than brands they do not. Danny Goodwin's write-up puts the gap at 3.2 times: models searched for familiar brands 55.7% of the time, against 17.4% for brands sitting outside a model's top 10.
The study covered 3,960 responses to 66 US buyer questions. It is a measurement of what models choose to look up, which is a narrower thing than rankings or traffic, and worth reading precisely for that reason.
What the study measured
Search Engine Land reported on 30 July 2026 that geoSurge measured how often AI models ran a search for a given brand, then compared brands the model already knew against brands it did not. Familiar brands were searched for 55.7% of the time. Brands outside the model's top 10 were searched for 17.4% of the time. The researchers also measured model memory and model search behaviour separately, so they could test whether what a model already knew shaped what it went on to look up.
The gap holds across every industry tested
The spread by industry was wide, Search Engine Land reported on 30 July 2026. Familiar brands were searched for between 41% and 82% of the time depending on the industry, and unfamiliar brands between 9% and 23%. The two ranges do not overlap. Even the industry where familiar brands did worst, at 41%, sat above the industry where unfamiliar brands did best, at 23%.
Familiarity concentrates on about five brands
When a model did search for a specific brand, 63% of those searches involved one of its five most familiar brands, according to the geoSurge figures reported by Search Engine Land on 30 July 2026. That is the finding with the sharpest edge. Being known to a model is not a wide club with room at the back. Most of the brand-specific searching pooled into a handful of names per topic.
Most searches name no company at all
Only 31% of fan-out searches included a company name, the same 30 July 2026 report said. Roughly two in three searches a model fired off were about the category, the problem or the specification rather than any brand. Category-level content therefore sits in front of the larger share of that activity, even though the brand-name minority is where the concentration effect bites.
Who ran the study, and what that means for the numbers
geoSurge is a vendor in the AI-visibility space publishing research about AI visibility. The figures are its own, and Search Engine Land reported them rather than independently replicating them. That does not make the numbers wrong, and it does mean they arrive from a party with an interest in the conclusion. Read them as a vendor study reported by trade press, which is what they are.
What the study does not show
The study measured which brands models search for. It did not measure rankings, clicks, traffic or revenue, and none of those can be derived from it. The questions were US buyer questions, and no Thai, APAC or non-English sample was reported. A model choosing to look up a brand is a step in a chain, not the end of one.
The numbers side by side
Every figure below comes from the geoSurge study as reported by Search Engine Land on 30 July 2026.
| What was measured | Reported result |
|---|---|
| How often models searched for familiar brands | 55.7% of the time |
| How often models searched for brands outside the top 10 | 17.4% of the time |
| Gap between the two | Familiar brands searched 3.2 times more often |
| Range by industry, familiar brands | 41% to 82% |
| Range by industry, unfamiliar brands | 9% to 23% |
| Brand searches going to a top-five familiar brand | 63% |
| Fan-out searches naming any company | 31% |
What this means for Thai marketers
The sample was 66 US buyer questions, so nothing here describes how models behave on Thai-language questions or in Thai categories. Treating the percentages as local numbers would be an invention. The mechanism, though, is the part worth thinking about, and mechanisms travel further than percentages do.
If prior familiarity shapes what a model looks up, then the off-platform work that builds familiarity carries weight it did not obviously carry before: consistent naming, coverage in places models read, and being described the same way across the sites that mention you. For a Thai brand competing against international names in the same category, the concentration finding is the uncomfortable one, because a handful of well-known brands can absorb most of the brand-specific searching. The practical response is measurement first. Check whether models name you at all for your category questions, which is the starting point of any GEO programme, then work on how your pages answer the category questions that make up the larger share of fan-out searches, which is where AI SEO and ChatGPT visibility work overlap.
Frequently asked questions
Does being familiar to an AI model mean more traffic?
The study does not show that. It measured how often models searched for a brand, not clicks, rankings, sessions or revenue, and no traffic figures were reported. Search behaviour is upstream of traffic, and the distance between them was not measured here.
Were Thai or other Asian markets included?
No. The sample was 3,960 responses to 66 US buyer questions, and no Thai, APAC or non-English sample was reported. The source does not say how models behave on Thai-language questions.
Who produced this research?
geoSurge, a vendor in the AI-visibility space, produced the study, and Search Engine Land reported it on 30 July 2026. Search Engine Land did not replicate the work independently, so the figures are the vendor's own.
Do I have to do anything about this today?
Nothing in the study requires an immediate change, because it is research rather than a platform announcement. The reasonable first step is to find out whether AI models mention your brand for your own category questions, which is a measurement you can take without changing anything.
If you want to know whether AI models name your brand when buyers ask about your category in Thailand, our team can run that check and show you what comes back. Get in touch and we will start with the questions your buyers actually ask.
