Start where the buyer starts.
An open category question reveals what a system brings to mind before a brand name steers the answer.
THE NEXT LAYER OF BRAND INTELLIGENCE
When consumers ask AI what to buy,
which brands get recommended?
AI Brand Explorer studies the answers. We repeat real consumer questions across AI systems to measure which brands appear, which get recommended, and how much the answer varies.
Ask more than once.
A recommendation is an observation. Repeated, independent answers let us study the distribution behind it.
Show what the data supports.
Report the model, the prompt, the sample, and the uncertainty. Keep a mention separate from a recommendation.
FROM THE RESEARCH
Measured API responsesA brand can remain in the evidence.
And disappear from the answer.
In one controlled study, marking a retailer unreachable changed brand inclusion even when the same product listings remained visible.
400 responses across four categories · Muse Spark 1.3 · 27 September 2026. The chart shows the carry-on cell, 25 responses per condition.
Read the study and its limits →Samsonite brand inclusion
CATEGORY BENCHMARKS / IN PREPARATION
The mattress shelf,
as AI sees it.
Our next study begins without a brand, a budget, or a sleeping preference. It asks how consistently each system recommends a leading brand, and how often a small sample could point to the wrong leader.
View the study design →Proposed baseline designNo results yet
50observations planned
per model
per model
6proposed Consumer AI
family proxies
family proxies
Empty slots illustrate the study plan. They are not responses or results. Exact API and consumer-app equivalence remains subject to review.
RESEARCH LIBRARY
Evidence, with the method attached.
Retail recommendationsCompleted study
When access changes the recommendation
A controlled test separates product evidence from an instruction about retailer access.
Frozen fixtures and three required recommendations per answer. Tests an access-status instruction, not actual browsing, real retailer blocks, or a free consumer app.
MattressesProtocol drafted
What’s the best mattress to buy?
A planned study of recommendation probability and how much a small sample can miss.
Findings have not been released. No brand ranking, recommendation share, or reliability result is available.
WHAT DOES “AVAILABLE” MEAN?
A brand’s chance of entering
the answer is a question we can study.
We call this statistical availability: the observed frequency of a defined brand outcome under a specific prompt, model, and collection protocol. It is not a universal brand score.
Read our measurement approach →