127 ChatGPT answers to 64 buyer questions in 5 EU niches: who gets linked
We asked ChatGPT 64 buyer questions about tea, coffee, mattresses, supplements and beauty supplements in France, Belgium and Ireland, twice each (127 answers, 13 Oct 2026 to 14 Oct 2026). The most-linked shop appears in 38% to 88% of answers, and the top three take 26% to 56% of all shop links.
8 niche-and-country cells · 127 answers · ChatGPT, logged out · Download CSV · One-page PDF
The eight cells
| Cell (full table on its page) | Answers | Most-linked shop | Top-3 share | Sites seen | Run overlap |
|---|---|---|---|---|---|
| Thé en vrac, France | 16 | palaisdesthes.com (10/16) | 31% | 41 | 35% |
| Café de spécialité, France | 16 | lomi.coffee (6/16) | 26% | 40 | 28% |
| Café de spécialité, Belgique (FR) | 16 | orcoffee.be (12/16) | 43% | 29 | 51% |
| Specialty koffie, België (NL) | 16 | orcoffee.be (10/16) | 38% | 34 | 44% |
| Matelas, France | 16 | tediber.com (14/16) | 56% | 19 | 61% |
| Compléments alimentaires, France | 16 | nutripure.fr (11/16) | 46% | 26 | 41% |
| Compléments beauté, France | 16 | nutriandco.com (6/16) | 31% | 26 | 40% |
| Specialty coffee, Ireland | 15 | cloudpickercoffee.ie (8/15) | 32% | 26 | 54% |
Top-3 share: the three most-linked shop sites as a share of all shop links in the cell. Run overlap: average share of shop sites that both runs of the same question linked.
What the numbers say
- MeasuredLinking is concentrated. The most-linked shop appears in 38% to 88% of a cell's answers; the top three shops take 26% to 56% of shop links.
- MeasuredA single answer is a poor guide. Two runs of the same question share only 28% to 61% of the shop sites they link, on average per cell.
- MeasuredLinks go to shops' own sites: 476 of 506 links across the eight cells (94%). Department stores, maps and review sites take the rest.
- MeasuredTechnical hygiene does not separate linked shops from unlinked ones. In the four cells with a comparison group, 24 most-linked shop sites we could check against 68 shops never linked: Organization schema 13 of 24 vs 44 of 68; Product schema 15 of 24 vs 33 of 68; OpenAI's search crawler not blocked 24 of 24 vs 65 of 68; page text in the raw HTML 22 of 24 vs 60 of 68. Small samples; they show only that these traits do not split the two groups.
- MeasuredChatGPT linked 13 domains with no working page, in 5 of the 8 cells, 22 times in total. They are excluded from the rankings. One of them was linked six times in a single cell.
- Not tested yetWhether published facts (starting price, delivery time, subscription terms), reviews or third-party mentions explain who gets linked. That needs a linked-versus-unlinked comparison on those traits; it is the next measurement round.
Method and limits
- Engine: ChatGPT (chatgpt.com), logged out, web search on, temporary chat, connection from Spain.
- Questions: 64 buyer questions in French, Dutch or English (8 per cell), each asked twice in fresh chats. 127 complete answers kept; failed runs are dropped.
- “Linked”: the shop's domain is among the answer's links. We do not count names in running text.
- Coverage: 5 niches, 3 countries, 3 languages. 127 answers is a small sample: read the percentages as counts, not as market share.
- Limits: one engine, one run location, runs from 13 Oct 2026 to 14 Oct 2026; answers change from day to day. No traffic or sales estimates. Domains are as ChatGPT linked them; we checked on 14 Oct 2026 which ones load, and ran the technical checks that day.
- Raw data: every answer is stored with its timestamp. Pooled CSV: download.
Is your shop in these tables?
We run the same kind of questions for your category and count how often your shop is linked. Free, one page, in your language.