Ask an AI chatbot where to buy a brand's products and you might get a polished, confident answer that points straight at a fake shop. A new study by intellectual-property protection firm Podqi, first reported by The Verge on November 12, found that ChatGPT, Google Gemini and Perplexity all directed shoppers to risky retail sites — including fake storefronts, counterfeit sellers and abandoned domains re-registered by third parties.
Podqi ran shopping questions about 151 brands across two testing periods in September 2026, producing 6,795 answers in total. Risky sites appeared in about 9 percent of all answers. Perplexity fared worst, surfacing a risky site in 15 percent of its answers, followed by ChatGPT at 8 percent and Gemini at 4 percent. In total, the researchers identified 718 risky domains: 351 fake sites and 367 inactive ones that could be repurposed for fraud.
Smaller brands hit hardest
Queries about smaller brands surfaced risky sites about twice as often as queries about big names. The brands tested included Lululemon, Birkenstock, Le Creuset and Stüssy. In one documented example, Perplexity named an incorrect page as Stüssy's official website and described it as the best source for authentic products.
The study counted recommendations, bare links and warnings together, so it measured how often risky URLs appear in answers rather than how often a chatbot actively vouched for a dodgy site. It also did not track whether anyone clicked through, bought anything or lost money. But the exposure alone matters: a shopper who follows an AI-generated link to an impersonator risks handing over payment details, login credentials or personal information to a fraudulent operator, and could end up with counterfeit goods or nothing at all.
Warnings are rare
ChatGPT warned users in only 24 percent of the instances where it surfaced a risky site — the highest warning rate of the three, which says plenty about how rarely any of them flag the danger. A Perplexity spokesperson, Jesse Dwyer, said the company “never claim[s] to be 100% accurate.” Google stressed that Gemini recorded the lowest exposure rate in the study.
Check the address bar before you pay
Podqi chief executive Trevor West advised shoppers to verify the source before entering personal or payment information. The practical takeaway is simple: type the brand's web address yourself or use its official app, check the domain carefully, and treat any chatbot shopping link — however confident it sounds — as a suggestion to verify, not a recommendation to trust.
The findings land as AI shopping assistants move from answering questions to completing purchases on users' behalf, with retailers and platforms racing to put chatbots at the centre of online shopping. The study is a reminder that the confident tone of an AI answer is not evidence that the link behind it is safe.
