Who gets named?
We scanned 84 well-known brands across six sectors, then asked open AI models the questions their buyers ask. Here is who is ready, and who gets named.
The short version.
Six things the data says.
- 01
Most sites are easy to read. Few answer questions.
The average site scored 76 out of 100. Indexing, crawler access and a clear next step are largely solved. Questions covered averaged 32, and only 22% of sites have three or more question-style headings or FAQ items.
- 02
AI models named the average brand in 42% of answers.
That is among well-known brands. 9 of the 40 brands we measured were not named once in 12 answers to their own buyer questions.
- 03
A better site helps less than you would expect.
Readiness and AI mentions moved together only weakly (correlation 0.24). This sample cannot say why, but it is a reason to work on what other sites say about you as well as on your own pages.
- 04
Which model you ask changes the answer.
On the same questions, Mistral Small named the brand 53% of the time, Llama 3.3 46% and gpt-oss 25%. One assistant is not a fair test.
- 05
B2B and media sites are the most closed off.
They scored lowest overall (69) and lowest on AI readiness (GEO 54). All 4 sites that block an AI search crawler are in this group.
- 06
llms.txt is spreading, mostly by default.
60% of sites answered at /llms.txt. Among Shopify stores it was 85%, against 44% of everyone else, which suggests the platform is doing much of the work.
Readiness by sector.
Average scores out of 100 for the homepage of each site.
| Sector | Overall | SEO | AEO | GEO | CRO |
|---|---|---|---|---|---|
| Clinics & wellness11 sites | 79 | 83 | 74 | 76 | 84 |
| Consumer products12 sites | 78 | 79 | 64 | 75 | 96 |
| Local businesses10 sites | 78 | 85 | 69 | 75 | 83 |
| Fashion & beauty12 sites | 76 | 78 | 63 | 72 | 92 |
| Food & drink13 sites | 75 | 81 | 61 | 66 | 91 |
| B2B, software & media14 sites | 69 | 81 | 68 | 54 | 72 |
How often AI names you.
The share of AI answers that named the brand, on questions its own buyers ask.
Bars show the average share of 12 AI answers naming each brand. The thin line is the 90% range. Ranges overlap, so read small gaps between sectors as noise.
Does a better site get named more?
Slightly. Each dot is one brand, and the correlation is 0.24.
Share of AI answers naming the brand
Each dot is one brand. Further right means a more ready site, higher up means named more often. The dotted line is the trend.
6 of 40 brands had below-median readiness and were still named in at least half of their answers.
7 of 40 had above-median readiness and were named in fewer than half.
What sites get right and wrong.
Average score for each signal, weakest first.
Models disagree.
Share of answers naming the brand, same questions.
By platform.
Average readiness, platforms with five or more sites.
What to do about it.
Four moves, in the order we would make them.
The weakest signal in the study, averaging 32 out of 100. Add a real FAQ, written in the words people type.
Read the AEO guide →State your facts as plain sentencesOnly 24% of sites give three or more kinds of quotable fact: prices, hours, places, phone numbers or results.
Read the GEO guide →Check who you are letting in4 sites block an AI search crawler, and 6 block one used for training. Make sure yours is a choice, not an accident.
See how to check →Earn mentions off your own siteReadiness alone did not decide who got named. Reviews, listings and articles on other sites are what assistants read too.
How GEO works →How we did this.
Read this before quoting a number.
- What we scanned
- 84 sites chosen by hand to be recognisable names in six sectors, none of them popsoda clients. 9 blocked our scanner and 3 returned almost no text, which left 72. Each was judged on its homepage across 12 signals.
- How AI answers were collected
- For 40 brands, Llama 3.3 wrote four buyer questions that did not name the brand. Llama 3.3, Mistral Small and gpt-oss each answered them, and we counted how often the brand appeared.
- Who is left out
- Brands whose names are everyday words are not in the AI counts, because a text match cannot tell them apart. B2B, software & media brands were not part of this AI run.
- What this cannot tell you
- These are open models, not ChatGPT, Gemini or Claude, so results may differ on those assistants. With 10 to 14 sites per sector, small differences are noise. Well-known brands probably do better than most small businesses.
- Why there are no names
- We publish sector-level results only. This is a snapshot of one page on one day, not a verdict on any company.
See the numbers as a table
| Sector | Sites | Overall | SEO | AEO | GEO | CRO | AI answers naming the brand | 90% range | Brands in AI run |
|---|---|---|---|---|---|---|---|---|---|
| Fashion & beauty | 12 | 76 | 78 | 63 | 72 | 92 | 29% | 15 to 46% | 9 |
| Consumer products | 12 | 78 | 79 | 64 | 75 | 96 | 57% | 32 to 80% | 7 |
| Food & drink | 13 | 75 | 81 | 61 | 66 | 91 | 45% | 30 to 59% | 11 |
| Clinics & wellness | 11 | 79 | 83 | 74 | 76 | 84 | 35% | 20 to 49% | 7 |
| Local businesses | 10 | 78 | 85 | 69 | 75 | 83 | 43% | 25 to 60% | 6 |
| B2B, software & media | 14 | 69 | 81 | 68 | 54 | 72 | not measured | 0 |
See how youscore.
Add your site or store, then a competitor, and see where you stand.