We asked AI about 20 small US brands. 18 of them never came up.

⏰ The short version

• We ran 20 indie US DTC brands through the 30 questions their customers actually ask AI.
• 18 of 20 (90%) came up in zero of them. Average: 0.10 out of 30.
• These weren't bad brands. Bean-to-bar chocolate makers, one-woman ceramics studios, herbalists with 30 years in.
• We ran the same test on 284 Korean brands and got 65% at zero. Different market, same gap.
• Nobody has claimed this ground yet. That's the whole point.

Why we counted

"AI search matters" is easy to say. What's hard to find is a number for how badly brands are actually missing. So we measured it.

The method is simple. Crawl the brand's site, pull the body text, then check it against the 30 questions a shopper in that category would type into ChatGPT. For each question: is there something here that could be quoted as an answer?

💡 What "counts as an answer" means Not whether a keyword appears. Whether a sentence exists that could be lifted into an AI answer. AI can't cite "Radiance Serum ✨". It can cite "vitamin C serum for sensitive skin, fragrance-free."

The result: 90% at zero

Questions answered	Brands
0	18
1	2
2+	0

Two brands scored 1. Everyone else scored nothing.

Here's what that means in practice. When a shopper asks AI "best natural dog treats without fillers?", AI names two or three brands. If you're not one of them, you don't exist for that shopper. There's no page two in an AI answer.

The uncomfortable part is that this leaves no trace. No impression, no click, nothing in your analytics. You don't lose visibly. You just never show up.

These were good brands

The list included:

• An Appalachian herbalist making hand-poured tinctures since the 90s
• A Hawaii bean-to-bar maker growing its own cacao on a North Shore farm
• A one-woman ceramics studio in Oakland throwing every piece by hand
• A chef-owned dog treat company upcycling rescued food from family farms

Every one of them has a real story and real craft. None of them come up in AI answers.

That's the finding. The gap isn't quality or authenticity. It's that the story isn't written in a shape AI can quote.

Korea showed the same thing

We'd already run this on 284 Korean DTC brands. 65.5% scored zero there.

So the obvious question was whether Korea was just behind — GEO is an English-language concept, most of the material is in English, maybe US brands were already on it.

They weren't. If anything the small US brands we tested scored worse.

💡 On sample size 20 brands is not a representative sample of US commerce. Don't read this as "the US is worse than Korea." Read it as: the US isn't prepared either.

Why an empty shelf is good news

Usually "we're late" means the seats are taken. This data says the opposite.

AI answers name two or three brands. The shelf is short — which cuts both ways. It's hard to get on, but whoever gets on first tends to stay, and right now most of those slots are unclaimed in both markets.

Getting your answers in place before the competition notices is the cheapest land grab available right now.

Honest footnotes

• 20 brands is a small sample. Directional only, not statistically representative.
• The US test used 30 questions; the Korean test used 50. Zero-rates are comparable; the averages are not.
• We excluded large, well-funded brands. Those may well score differently.
• Data as of July 2026.