AI chatbots surface risky shopping sites, study finds

AI chatbots can send shoppers to fake stores, counterfeit sellers and dead websites, a new study has found. The research, by intellectual property protection firm Podqi and first reported by The Verge on 12 November, found risky retail sites in about 9 per cent of 6,795 answers from ChatGPT, Google Gemini and Perplexity, Gadget Review reported.
Perplexity performed worst, surfacing a risky site in 15 per cent of its answers, followed by ChatGPT at 8 per cent and Gemini at 4 per cent. Podqi asked the chatbots shopping questions about 151 brands, including Lululemon, Birkenstock, Stussy and Le Creuset, in two rounds in September.
718 risky domains found
Podqi reviewed 4,437 domains and flagged 718 as risky: 351 fake sites and 367 inactive ones. A site counted as risky if it impersonated a real retailer, sold counterfeit goods, was an old brand domain re-registered by someone else, or was a dead domain a fraudster could reuse. In one example, Perplexity named the wrong page as the official Stussy site and described it as a source of authentic products.
The chatbots rarely warned users: in 57 per cent of answers containing a risky site, the chatbot actively recommended it, and risky sites were recommended 6.3 times as often as warned against. ChatGPT warned in only 24 per cent of cases where it surfaced one, the highest rate of the three.
Check before you pay
Podqi chief executive Trevor West advised shoppers to verify any source before entering payment details and watch for misspelled domains. The findings land just before Black Friday on 27 November, when millions search for deals. The study has limits: Podqi wrote its own prompts rather than analysing real conversations, and it did not track whether anyone clicked a risky link or lost money, Digit reported.