We asked ChatGPT 360 shopping questions about home goods. Here is who it recommended and where it looked.
When a shopper asks ChatGPT "what are the best linen sheets?" or "a good sectional under $3,000?", it gives a short list of brands. We wanted to know two things: how often does a mid-size home brand make that list, and which web pages does ChatGPT read before it answers?
So we took 36 US home-goods brands that sell on Shopify (furniture, bedding, tableware, candles, decor) and asked 10 shopper questions for each one. The questions came from each store's own catalog: its product types, its price range, the occasions people buy for. That gave 360 questions. ChatGPT answered 355 of them.
The short version
- A brand was named in 12% of the answers about its own category (43 of 355).
- 21 of the 36 brands were never named, in any of their 10 questions.
- ChatGPT cited web pages in 92% of answers, about 8 pages per answer.
- It read the brand's own website in only 6% of answers. In every one of those answers, it named the brand.
- 4 in 10 answers leaned on a "best of" list or review. The brand was missing from 85% of the lists we checked.
1. Most brands are not in the answer at all
| How often the brand was named | Brands |
|---|---|
| Never (0 of 10 questions) | 21 |
| Once or twice | 9 |
| Three times or more | 6 |
These are established stores with full catalogs. The questions were about products they actually sell, at prices they actually charge.
The names that did come up were often the same few. Brooklinen alone appeared in at least 18 answers across the bedding questions.
2. ChatGPT mostly reads other people's pages
We sorted the 2,075 different pages ChatGPT cited:
| Type of page | Share of cited pages |
|---|---|
| Another brand's own store | 38% |
| Retailers, resale and other sites | 37% |
| "Best of" lists and reviews | 12% |
| Marketplaces | 9% |
| Magazines and news | 2% |
| Video and social | 1% |
| The store the question was about | 0.3% |
The single most-cited sites were not brand sites at all:
| Site | Answers that cited it |
|---|---|
| eBay | 33 |
| Home Depot | 29 |
| Etsy | 28 |
| Homes & Gardens | 26 |
| Walmart | 24 |
| Sleep Foundation | 22 |
| Macy's | 22 |
| Poshmark | 20 |
| Lumens | 20 |
| 19 |
A brand that sells only through its own site is competing with every page on these sites that mentions a similar product.
3. When ChatGPT reads your site, it names you
ChatGPT cited the brand's own website in 23 answers. It named the brand in all 23.
In the other 332 answers it did not open the brand's site, and it named the brand in only 20 of them.
This is a pattern, not proof of cause. A brand ChatGPT already knows is more likely to be both read and named. But it matches what the page types show: the answer is built from the pages it reads, and most of the time those pages belong to someone else.
4. "Best of" lists carry a lot of weight, and most brands are not on them
39% of answers cited at least one "best of" list or review page. Think "14 best sectional sofas under $3,000" or "best linen sheets, tested".
We opened 65 of those lists and looked for the brand the question was about. It was on 10 of them. It was missing from 55.
When a list is the source and you are not on the list, you are not in the answer.
What a brand can do with this
- Ask your own questions. Type the five questions your shoppers really ask into ChatGPT and write down who it names and which pages it links. That is your real competition, and it is often not who you think.
- Find the lists. For each linked "best of" page, check whether you are on it. If not, that page is worth one honest email to the writer.
- Put the facts on the page. Size, material, weight, where it is made, who it is for. Written as plain text on the product page, not hidden in an image.
- Check again later. Answers change from week to week. One check is a snapshot; a trend needs the same questions asked regularly.
How we did it, and the limits
- Brands: 36 US home-goods brands on Shopify. We picked them as typical mid-size stores; this is not a random sample of all brands.
- Questions: 10 per brand, written from the store's public catalog. 360 in total; 355 got an answer.
- Model: OpenAI's API (GPT-4.1 mini with web search, US location), one answer per question. API answers are close to what people see in the ChatGPT app, not identical.
- "Named" means the brand's name appeared in the answer.
- One run per question. ChatGPT's answers vary from run to run, so single numbers here have a wide margin. The overall pattern across 355 answers is steadier than any one brand's score.
- We do not publish the names of the brands we checked.
Pickory checks every week what ChatGPT, Claude and Perplexity recommend for a Shopify store's own questions. A free one-time check for your store: getpickory.com.