A week before Black Friday, two BBC technology reporters asked four AI assistants to find the best UK deal on a 55-inch TV. What came back were sold-out listings, dead links and deal pages from last year. I build selling agents for webshops, and this test shows the part of AI shopping that worries me most for small shops. The assistants repeat whatever your store left lying around.
Liv McMahon and Imran Rahman-Jones used the free versions of ChatGPT, Claude, Gemini and Amazon's Alexa with the same prompt. They asked for the best UK Black Friday deal on a 55-inch TV, with a budget of £500 to £700, good picture quality and regular gaming, and asked each assistant to compare real retailer prices. They said themselves that it was not a scientific test.
ChatGPT's top pick was a Hisense TV on Joybuy at £499, down from £649. Its safer option was the same TV at full price from John Lewis and Richer Sounds, but when the reporters clicked through, neither shop still listed it online.
Claude recommended two TCL TVs. Its links went to dead pages on the Richer Sounds website, and some other links went to round-ups of Black Friday 2025.
Gemini gave no links at all. It listed one TCL model at £539 at Argos and Currys. The Argos price was right. At Currys the same TV cost £60 more.
Alexa got every price right, but it only searched Amazon, and it told the reporters to check other shops.
Naureen Mohammed, who helps businesses appear in AI results, told the BBC that when shops do not update their prices, "last year's prices are served up as today's deals." Her own analysis of chatbot answers about UK Black Friday tech deals found Amazon in half of them, with Currys close behind.
The assistants found those dead pages and old deals on real retailer sites. Old product URLs, sold-out listings and last year's sale pages are all part of what an AI reads about your shop.
A small skincare shop faces the same risk. Think of a discontinued serum page that still loads, a gift set page with last December's price, or a shade you stopped selling. When an assistant sends a shopper to that page, the shopper sees your shop name next to the wrong price and leaves.
The one assistant with correct prices, Alexa, only searched one catalog, its own. And in Naureen Mohammed's analysis the big retailers showed up most often. A small shop does not get the benefit of the doubt, so every page an assistant can find needs to be current.
The test also shows a second gap. The prompt gave a budget, a screen size and a use. The answers still included a 50-inch TV, a TV £150 over budget and a paid ad for an OLED model at £1,100. Skincare shoppers give the same kind of details: skin type, budget, what they use now. An answer that ignores them sends people to the wrong product.
Black Friday is on 27 November. Do these first.
That last point is what we build at CAIDEL. A selling agent on your WooCommerce store reads your current catalog, so the price it quotes is today's price. It asks the questions a good advisor asks before it recommends, such as skin type, budget and current routine. And you get the record of what buyers asked, which shows you where your product data has gaps.
Want to see how AI agents read your store today? We built a free snapshot for WooCommerce shops.
Why do AI assistants show old prices for my products?
They answer from pages they can find, including old product pages, past sale pages and deal round-ups. If your old pages still load with old prices, an assistant can present them as current deals. Redirect retired pages and put sale prices on a schedule.
Should a small WooCommerce shop worry that big retailers show up most in AI answers?
In the BBC test and in Naureen Mohammed's analysis, big retailers appeared most often. A small shop cannot outspend them, but it can make sure every page an assistant finds is current, and it can answer shoppers well on its own site.
How can I check what AI assistants say about my shop?
Ask ChatGPT, Gemini and Claude the questions your customers ask, with your category and a budget. Click every link they return, note wrong prices, dead pages and missing products, and fix those pages first.