AI Shopper Index

Which brands do AI assistants actually recommend?

We ask ChatGPT, Claude, Perplexity and Gemini what to buy in a brand’s own category, then count how often that brand is named.159 brands ranked across 4 assistants. Last full reading: 14 Sept 2026.

What changed

Share of category answers, against each brand’s previous reading. 92 brands had a reading to compare, 55 held steady and 65 are on their first run.

Everlane fashion2.9 percentage points down since the previous reading
Gymshark fashion2.2 percentage points up since the previous reading
Ministry of Supply fashion1.7 percentage points up since the previous reading
Filson fashion1.3 percentage points down since the previous reading
Smartwool fashion1.2 percentage points down since the previous reading
Faherty fashion1.0 percentage points up since the previous reading
Darn Tough fashion0.7 percentage points up since the previous reading
Alo Yoga fashion0.7 percentage points up since the previous reading

The ranking

183 of 183 brands ranked

CategorySortSearch

Each brand is scored against its own category’s pool of assistant answers, so a share is comparable within a category and not across them. 66 of 183 brands shown were named at least once.

Rank	Brand	Share of voice (ranked)	Change	Named in	Recommended first	Products named	Agent readability	Can agent buy	Tracking	Landing page
	Aeraersf.comAccessories	Not yet scored					79	Yes2026-09	97	94
	Orbitkeyorbitkey.comAccessories	Not yet scored					69	Yes2026-09	94	93
	Peak Designpeakdesign.comAccessories	Not yet scored					48	Yes2026-09	14	88
	qwstionqwstion.comAccessories	Not yet scored					Not measured	Not checked	Not measured	Not measured
	Ridgeridge.comAccessories	Not scored: the store blocks automated readers					Not measured	Not checked	Not measured	Not measured
Electronics3 brands against 253 answersNo brand here was named in any of these answers, so there is no rank 1 to show.
	Courantstaycourant.comElectronics	0%under 3.0%	No change	0 of 253			80	Yes2026-09	97	75
	Incaseincase.comElectronics	0%under 3.0%	0.4 percentage points down since the previous run	0 of 253			74	Yes2026-09	94	92
	Twelve Southtwelvesouth.comElectronics	0%under 3.0%	No change	0 of 253			78	Yes2026-09	94	85
Fashion89 brands against 416 answers
1	Everlaneeverlane.comFashion	9.4%	2.9 percentage points down since the previous run	39 of 416	0.96%	14	73	Yes2026-09	94	83
2	Darn Toughdarntough.comFashion	7.2%	0.7 percentage points up since the previous run	30 of 416	4.1%	13	86	Partly2026-09	97	87
3	Smartwoolsmartwool.comFashion	6.7%	1.2 percentage points down since the previous run	28 of 416	1.7%	15	Not measured	Not checked	Not measured	Not measured
4	Allbirdsallbirds.comFashion	5.0%	No change	21 of 416	2.9%	10	80	Partly2026-09	97	93
5	Ministry of Supplyministryofsupply.comFashion	3.4%	1.7 percentage points up since the previous run	14 of 416	0.48%	2	53	No2026-09	100	77
6	Gymsharkgymshark.comFashion	2.9%	2.2 percentage points up since the previous run	12 of 416	0%	8	18	Partly2026-09	94	52
7	Ten Thousandtenthousand.ccFashion	2.4%	0.4 percentage points up since the previous run	10 of 416	0.96%	7	80	Yes2026-09	94	94
8	Organic Basicsorganicbasics.comFashion	2.2%	0.5 percentage points up since the previous run	9 of 416	0.24%		91	Yes2026-09	88	75
9	Fahertyfahertybrand.comFashion	1.9%	1.0 percentage points up since the previous run	8 of 416	0.48%	4	99	Yes2026-09	97	73
10	Outdoor Voicesoutdoorvoices.comFashion	1.7% (within noise at this sample)	0.1 percentage points down since the previous run	7 of 416	0%	4	78	Yes2026-09	97	88
10	Outerknownouterknown.comFashion	1.7% (within noise at this sample)	0.1 percentage points down since the previous run	7 of 416	0.24%		93	Yes2026-09	85	81
12	Alo Yogaaloyoga.comFashion	1.2% (within noise at this sample)	0.7 percentage points up since the previous run	5 of 416	0%	1	50	Not checked	97	64

Share of voice:: the share of the category’s answers that named the brand, 0 to 100%.
Recommended first:: of the answers that named the brand, how often it was the first pick.
Agent readability, Tracking, Landing page:: our own checks of the store, each 0 to 100: 80 and above is good, 50 to 79 fair, below 50 poor.
Can agent buy:: whether an agent could complete a purchase in our check of the store, with the month it was checked.

Within noise at this sample: the rate is real but below what this pool size can tell apart from zero.

Each category has one fixed set of unbranded shopping questions, the same for every brand in it, asked to all four assistants. A category with no full reading of those questions yet shows its brands as Not yet scored. A brand counts once in an answer that names it anywhere, in any spelling of its name; a brand whose name is an ordinary word or phrase counts only when its name comes with its own products.

Ranked on share of voice within a category. The other columns are reported beside it, not added to it. A brand named zero times is shown with the upper bound its category’s sample supports, because a zero from a large sample is a measurement and hiding it would publish silence as absence.

The index has not been populated yet

183 brands are queued. The first results land shortly.

Run the audit on your own store

How this is measured

Every number on this page is a count of answers we collected. Nothing here is modelled, estimated or inferred.

A brand is counted as named only on a word boundary match against the raw answer. No model judges whether a mention happened.
A reading is published only when all four assistants answered, and a category needs at least 100 answers before any brand in it is ranked. Below that we say "Not enough answers yet" rather than report a number.
A brand named zero times is published with the upper bound its sample supports, not hidden.
Stores that could not be read are listed with the date and the response we actually got, and carry no score or change.

The current ranking rests on 1,882 assistant answers across 6 categories, covering 159 brands. Data as of 14 Sept 2026, the latest reading all four assistants answered. Each category has one fixed set of unbranded shopping questions, the same for every brand in it, asked to all four assistants. A category with no full reading of those questions yet shows its brands as Not yet scored. It is a small sample by design, and we do not publish a score a sample cannot support: a category below 100 answers shows “Not enough answers yet” instead.

The full specification, including the failure modes this method has been wrong in before, is in the index methodology.

Want your brand measured?

Submitting lists your store publicly once it passes our checks.

What we check, all from your public storefront

Answers a plain request
Publishes a readable product catalogue
Classifies into a category we rank
Supports at least six shopping questions for its own report (not used by the index)

The last one is what most stores fail. We build the questions from your own product types, tags and collections, and a store we cannot ask six specific questions about cannot carry a number we would stand behind.

Nothing here is permanent. It describes what we could read today, and the result records which check did not pass and what would change it.