We asked one AI model 20 high-intent beauty shopping questions, 20 times each. Then we checked something nobody else is measuring: does being recommended by AI have anything to do with how AI-ready your store actually is?
Research is only worth anything if you can see how it was made. So here is the whole setup before any findings. Everything below is computed from real captured responses. Nothing is estimated or projected.
A brand's share of all recommendations captured. The whole field sums to 100 percent.
In what percent of the 400 prompt-runs the brand appeared at least once.
Average rank in the answer when the brand appeared. Lower is better.
Top 20 brands by share of voice across all 4,000 recommendations. The last column is each brand's real AI Commerce Score™ from our index, where the store is something we measure. Hold that column in mind, because the next section is about it.
| # | Brand | Share™ | Freq™ | AI Commerce Score™ | |
|---|---|---|---|---|---|
| 1 | Neutrogena | 5.6% | 55.8% | 3.6 | 43AI Invisible |
| 2 | The Ordinary | 5.0% | 49.8% | 3.9 | 50Low |
| 3 | CeraVe | 4.8% | 48.0% | 3.9 | 54Low |
| 4 | La Roche-Posay off-index | 4.5% | 45.0% | 3.9 | —Off-index |
| 5 | Drunk Elephant | 4.0% | 40.3% | 5.6 | 87Highly |
| 6 | Clinique | 3.9% | 38.5% | 5.3 | 14AI Invisible |
| 7 | Tatcha | 3.6% | 36.0% | 7.6 | 71Moderately |
| 8 | Paula's Choice | 3.3% | 33.0% | 4.5 | 73Moderately |
| 9 | SkinCeuticals | 3.2% | 31.8% | 4.1 | 14AI Invisible |
| 10 | Burt's Bees | 2.7% | 27.0% | 6.4 | 34AI Invisible |
| 11 | Kiehl's | 2.6% | 25.5% | 5.6 | 14AI Invisible |
| 12 | Estee Lauder off-index | 2.0% | 20.3% | 4.1 | —Off-index |
| 13 | Aveeno | 1.7% | 17.3% | 6.7 | 49AI Invisible |
| 14 | Olay | 1.5% | 15.3% | 4.3 | 61Low |
| 15 | Murad off-index | 1.5% | 15.0% | 8.2 | —Off-index |
| 16 | COSRX | 1.4% | 14.0% | 6.4 | 56Low |
| 17 | Vichy off-index | 1.3% | 12.8% | 7.3 | —Off-index |
| 18 | Mario Badescu off-index | 1.2% | 11.8% | 8.9 | —Off-index |
| 19 | L'Oreal Paris off-index | 1.1% | 10.5% | 7.6 | —Off-index |
| 20 | Herbivore Botanicals | 1.0% | 9.8% | 5.0 | 49AI Invisible |
If AI recommended the stores that are easiest for AI to read, these two numbers would move together. They do not. Across the on-index brands, the correlation between Recommendation Frequency™ and AI Commerce Score™ is just r = 0.17 (n = 18). At this sample size that is statistically indistinguishable from zero, well below the threshold for significance. The honest read is not a weak link, it is no measurable relationship. Each dot is a brand.
The mirror image is Drunk Elephant, the best-built store in the set at 87 out of 100, recommended often but not the most. And the single most recommended brand overall, Neutrogena, sits at a middling 43. The average AI Commerce Score™ across all on-index recommended brands is only 45.3, barely above the database-wide average. Being famous is currently enough.
When buyers ask AI for "the best" of something, AI answers with brands, almost never with shops. Retailers like Amazon, Sephora, and Ulta accounted for just 14 of 4,000 recommendations. That is the whole thesis of Recommendation Intelligence™ in one number: the contest AI runs is between brands and their own stores, not between marketplaces.
Today AI recommends from memory. It reaches for the names it saw most during training, which is why decades-old drugstore brands win even when their stores are nearly unreadable to a machine. That is a temporary state. As AI shopping moves from recalling names to live retrieval and agents that browse, compare, and check out, the advantage shifts to the stores an agent can actually read, trust, and act on.
That is the gap this research exposes and the one the Recommendation Intelligence Framework™ is built to close: AI Readability, AI Understanding, AI Trust, Recommendation Intelligence, and Decision Confidence. The brands coasting on fame today are the ones with the most to lose when the model changes. The brands building readable, trustworthy, machine-legible stores now are the ones that keep the recommendation when memory stops being enough.