Over the past two weeks we ran free Feedlegible audits on 45 Shopify stores, most of them US home, bedding and lifestyle brands. For each store we asked ChatGPT, Gemini and Perplexity the two shopping questions its customers are most likely to ask — "best duvet covers", "best weighted blankets", "best kids bedding sets" — and recorded who each assistant recommended.
That is 61 different questions and 249 store-and-assistant answers. Here is what they showed.
The headline numbers
- 32 of 45 stores (71%) weren't named in a single answer for the products they sell.
- Across all 249 answers, the store we were checking was named 38 times — 15% of the time.
- One brand, Brooklinen, was recommended in the answers for 24 of the 45 stores. L.L.Bean (21), Boll & Branch (18), Parachute (16) and Quince (15) followed.
If you sell bedding in the US, the assistants have a short list, and most stores aren't on it.
A high readiness score didn't mean being recommended
Every audit also scores how ready a store is for AI shopping agents: can an agent read the products, prices, ratings and policies, and can it buy?
You might expect the best-prepared stores to be recommended most. They weren't:
| Readiness score | Stores | Named in zero answers | Share of answers that named the store |
|---|---|---|---|
| 90 and above | 28 | 21 (75%) | 10% |
| 80–89 | 12 | 9 (75%) | 13% |
| Below 80 | 5 | 2 (40%) | 52% |
The stores scoring below 80 were mostly large, well-known brands. They were recommended anyway, because the assistants already knew them.
So readiness is not what gets you into the answer. It's what lets an agent use you once you're there: read your product correctly, quote your rating, show the right price and send the shopper to checkout. Both matter. They're different problems.
Where the recommendations come from
In an earlier round of checks, when ChatGPT showed where its picks came from, the sources were editorial roundups — Sleep Foundation, GearJunkie, NBC Select and similar "best of the year" lists. The brands that appear in those lists are the brands that get named.
That matches how these assistants work. Asked a general question like "best duvet covers", they summarise what trusted sources say. A store's own product page rarely is that source.
What this means for your store
1. Find out where you stand. Ask the assistants the questions your customers ask, in their words, and note who is named. Do it more than once — answers change from day to day.
2. Find the sources behind the answers. Look at which sites the assistants cite for your category. Those are the publications, review sites and comparison pages worth pitching, sampling to, or earning a place on.
3. Make sure agents can read what you'd want quoted. In the same 45 stores:
- 44 of 57 stores we could check had no star rating in the structured data agents read, even when reviews showed on the page.
- 35 of 58 had thin product descriptions — under 200 characters for most products.
- 14 of 57 had no Product structured data on product pages at all.
When an assistant does consider your product, these gaps decide whether it can say "rated 4.8 from 2,300 reviews, made of 100% linen, $129" — or nothing at all.
4. Ask narrower questions. Generic "best X" questions favour the biggest brands. Shoppers increasingly ask specific ones: "best linen duvet cover for hot sleepers under $200". Your product data — materials, sizes, care, who it's for — is what lets an assistant match you to those.
How we measured
- 45 Shopify stores audited between 8 and 17 October 2026, mostly US home, bedding and lifestyle brands; results lean towards that category.
- Two questions per store, chosen from each store's own description and largest collections, asked once to ChatGPT, Gemini and Perplexity as a US shopper.
- A store counted as "named" if the assistant recommended it by name. Answers vary by day, wording and user, so treat a single check as a snapshot, not a ranking.
You can run the same check on your own store — free, in about a minute.