Websites AI cites most in Retail

Retail prompts are transactional. Shoppers rarely ask an AI to explain a concept: they want a shop, a product or some reassurance before paying. Most requests fall into four families.

• Where to buy: "where can I order a personalised birthday card online", "best site to print a photo book", "shop selling glass baby bottles".
• Trust checks: "is this online store legit", "reviews of [shop name]", "has anyone ordered from this site".
• Comparisons: "Etsy or Amazon for handmade gifts", "cheapest place to buy car wax".
• Rights and after-sales: "how do I return an item bought online", "what to do if my parcel never arrived".

Each family pulls a different type of source into the answer, which explains why retail citations are spread over so many kinds of websites.

Four types of websites share most retail citations.

• Marketplaces such as Amazon, Etsy or Cdiscount act as product catalogues. An AI cites them when it needs proof that a product exists, a price range or a large pool of customer reviews.
• Review platforms and consumer associations, Trustpilot first, along with groups like UFC-Que Choisir in France, answer the "can I trust this shop" questions.
• Public consumer information sites, for example the French economy ministry's consumer pages or service-public.fr, appear whenever a prompt touches withdrawal rights, legal warranties or delivery disputes.
• Independent shops form the long tail: a specialist retailer with a clear product page can be named directly when the prompt is specific enough.

Social networks (Facebook shop pages and groups), YouTube videos and classified ads sites such as Leboncoin complete the picture, mostly for second-hand goods or "how do I use it" questions.

In classic search results, a marketplace often outranks a small retailer on generic product queries. AI answers behave differently when the prompt is precise. Someone asking for "a personalised wooden birth announcement" or "a shop that engraves pet name tags" needs a page that matches the request closely, and a specialist catalogue often provides that match better than a marketplace listing.

The reverse holds for broad prompts. "Gift ideas for a 10-year-old" is usually answered from marketplaces, gift guides and media roundups. A retailer that only exists on its own website has little chance there unless it also appears in those third-party lists.

The assistant matters too. Those that search the web for every answer, like Perplexity, readily link to product pages and marketplaces, while a model answering from its training data tends to name brands it already knows.

Retail is a broad category. It gathers shops selling gifts, stationery, childcare products, car accessories and many other goods, so the sites at the top reflect the products covered by the tracked questions, not the market share of each retailer.

Most of the questions behind this dataset are asked in French, which is why French public sites and French retailers are so visible. A shop based in another country should focus on the types of sources cited rather than on individual names.