Which brands does Claude recommend? The Berakio AI Index, November 2026
When a shopper asks an AI assistant "what's the best matcha to buy online?", the answer names a handful of brands. Those names are the new shelf. If yours isn't on it, the shopper may never see you.
We wanted to know what that shelf looks like for small, independent brands. So we asked one assistant, Claude, the same shopping questions many times and counted which brands came back.
The work of getting into these answers has a name now: generative engine optimization, or GEO. But you can't improve a number you haven't measured. This index is that measurement.
How we measured
- Model: Claude Sonnet (Anthropic), with web search off. Each question went in as a fresh conversation with a plain "You are a helpful assistant" system prompt. This is close to, but not the same as, what someone sees in a consumer AI app.
- Questions: five ordinary shopping questions per category, asked word for word, with nothing added. One question per category in three categories was a "Where can I buy…?" question. Its answers mostly list stores, so we report it separately and leave it out of the brand numbers below.
- Repeats: each question was asked 10 times. The same question gets a different answer each time, so a single answer tells you very little. We report how often each brand appears across all the answers.
- Date: all answers were collected on 11 November 2026.
- Counting: a brand "appears" if the answer names it at least once. We built a name dictionary for each category to merge spellings ("Onyx" and "Onyx Coffee Lab" are one roaster), to keep stores and marketplaces apart from brands (Amazon, Chewy and YesStyle are stores; Amazon Basics is a brand), and to skip product names (a wallet model is not a brand).
- Categories: matcha, harnesses for small dogs, minimalist wallets, Korean sunscreen, and coffee beans.
All figures: Claude Sonnet (Anthropic), web search off, answers collected 11 November 2026.
| Category | Answers | Brands named | In at least half of answers | In two answers or fewer |
|---|---|---|---|---|
| Matcha | 40 | 24 | 6 | 9 |
| Small-dog harnesses | 40 | 16 | 6 | 3 |
| Minimalist wallets | 50 | 71 | 5 | 43 |
| Korean sunscreen | 40 | 27 | 8 | 12 |
| Coffee beans & roasters | 50 | 62 | 7 | 37 |
What we found
1. A small group of brands shows up almost every time
In every category, between 5 and 8 brands appeared in at least half of the answers. Some appeared in all of them. Across 50 minimalist-wallet answers, Bellroy and Secrid were named in every one.
2. Then there's a long tail
Behind that group, the list thins out fast. Across the five categories the answers named 200 different brands. 104 of them (52%) appeared in two answers or fewer. In wallets and coffee it was about six in ten.
For a small brand, this is where the fight is. If you're in the tail, a shopper who asks Claude once will probably not see you. A shopper who asks again might.
3. Four brands own Claude's answer on Korean sunscreen, and the old giants aren't among them
The most striking result came from Korean sunscreen. Four brands appeared in every one of the 40 answers: Beauty of Joseon, Isntree, Round Lab and Skin1004. None of them belongs to Korea's two largest beauty groups, Amorepacific and LG H&H. Amorepacific's Innisfree did well (60%), but its luxury brand Sulwhasoo was named in only one answer out of 40.
The four aren't all small independents. Beauty of Joseon, Skin1004 and Round Lab are now part of Goodai Global, a younger group that grew by buying K-beauty brands. Isntree, as far as public reports show, is still owned by its founder.
So it isn't the oldest or biggest parent company that wins this shelf, at least in Claude's answers. A brand from outside the two giants, and even one with no group behind it, can be in every answer Claude gives.
4. A marketplace's own label takes a seat
In minimalist wallets, Amazon Basics appeared in 68% of Claude's answers. That's more than every independent brand except Bellroy, Secrid and Ridge. It came up as the budget pick every time it appeared. The coffee answers show a similar pattern: Trade Coffee, a marketplace selling beans from many roasters, appeared in 58% of answers, and Costco's Kirkland Signature in 22%.
Independent brands aren't only competing with each other. They are also competing with the store's own label and with the middleman.
The top of each shelf
Share of Claude's answers that named each brand. Brand questions only. Claude Sonnet, web search off, answers collected 11 November 2026.
Matcha
40 answers.
| Brand | Share of answers |
|---|---|
| Encha | 97% |
| Jade Leaf | 97% |
| Ippodo | 90% |
| Marukyu Koyamaen | 75% |
| Aiya | 55% |
| Matchaful | 55% |
| Kettl | 33% |
| Naoki Matcha | 33% |
Small-dog harnesses
40 answers.
| Brand | Share of answers |
|---|---|
| Ruffwear | 97% |
| Puppia | 93% |
| 2 Hounds Design | 72% |
| Rabbitgoo | 72% |
| Kurgo | 65% |
| PetSafe | 57% |
| Julius-K9 | 47% |
| Blue-9 | 38% |
Minimalist wallets
50 answers.
| Brand | Share of answers |
|---|---|
| Bellroy | 100% |
| Secrid | 100% |
| Ridge | 86% |
| Amazon Basics | 68% |
| Dango | 52% |
| Zitahli | 48% |
| Saddleback Leather | 44% |
| Ekster | 42% |
Korean sunscreen
40 answers.
| Brand | Share of answers |
|---|---|
| Beauty of Joseon | 100% |
| Isntree | 100% |
| Round Lab | 100% |
| Skin1004 | 100% |
| Anua | 63% |
| Innisfree | 60% |
| Purito | 60% |
| Missha | 55% |
Coffee beans & roasters
50 answers.
| Brand | Share of answers |
|---|---|
| Counter Culture | 96% |
| Onyx Coffee Lab | 96% |
| Intelligentsia | 84% |
| Stumptown | 76% |
| Tim Wendelboe | 68% |
| Blue Bottle | 62% |
| Passenger Coffee | 56% |
| Peet's | 48% |
What this doesn't tell you
- One model, no web search. These are Claude's answers, not AI answers in general, and other assistants can answer very differently. When we put the same questions to another AI model in our own tests, it named far fewer brands in each answer, and in each category only a few brands were among the most-named in both. A brand that turns up in most Claude answers may rarely or never appear in another assistant's, and the reverse. Answers with web search on may differ again. We plan to add other assistants.
- One day. Model updates can change these numbers.
- Mentions are not endorsements. We count a name whenever it appears, including in phrases like "a cheaper alternative to…".
- The questions shape the answer. Different wording brings up different brands. We used plain wording and published every question.
How was this made?
The answers in this index were generated by an AI model (Claude Sonnet). The counting is done by a script and a hand-checked name dictionary. This write-up was drafted with AI assistance.
We don't sell placement, and no one can promise that an AI will recommend a particular brand. What we can do is measure.
Check your own brand
Curious how often AI names your brand when shoppers ask about your category? Run a free Berakio AI Check. Tell us your brand, your site and your category, and we'll ask the question 10 times and show you who comes up.
Tags: AI search visibility, GEO check, generative engine optimization, AI shopping recommendations, brand visibility in AI answers