Shoppers who turn to AI chatbots for product research or buying advice face a real risk of being steered toward counterfeit websites and other dangerous destinations, according to a study published by IP protection firm Podqi and shared exclusively with The Verge. The research tested ChatGPT, Gemini, and Perplexity across thousands of shopping queries and found that all three platforms directed users to risky sites, though at varying rates. Perplexity performed worst, displaying problematic websites in 15 percent of responses, while Gemini showed the strongest performance at just 4 percent.
The testing involved 6,795 answers generated across two periods in September, covering 151 brands ranging from Lululemon and Birkenstock to Le Creuset and Stüssy. Podqi created prompts asking where to buy specific brands, which sites sold them, and whether they shipped internationally with regional websites. ChatGPT landed in the middle, showing risky destinations in 8 percent of answers, but led the pack when it came to warning users about problematic sites, doing so 24 percent of the time those sites appeared. The study defined risky sites as fake storefronts impersonating legitimate retailers, knockoff sellers, expired brand domains that others had reregistered, or dead sites that could be bought and repurposed. Smaller brands faced roughly double the risk compared to larger companies, with queries about big-name retailers surfacing dangerous sites about half as often.
"Part of this piece, too, is making sure that the consumers know, even though you are being fed a result in a particular chatbot, it is always best to verify the source of that before entering personal information or credit card information," Podqi CEO and cofounder Trevor West said. The tested platforms sometimes fabricated fake regional sites for brands that don't operate any, and certain question types triggered risky results more frequently than others—particularly prompts about regional brand websites. In one instance, Perplexity incorrectly identified a page as Stüssy's official website and described it as "the best starting point for guaranteed authentic items."
The findings arrive as AI companies pitch their chatbots as useful tools for shopping tasks like researching products, comparing options, and discovering retailers—claims that could prove especially risky heading into the holiday shopping season, when consumers hunt for in-stock items and competitive prices. More shoppers are expected to use AI for purchasing this year, including buying from unfamiliar brands or websites. The problem creates headaches for brands and retailers trying to protect their reputations, but it poses a more immediate danger to consumers who might hand over payment information or personal details to fraudulent sites. OpenAI said its goal is "to connect people with reliable places to shop" and that it's working to improve safeguards against deceptive sites, while Google emphasized that Gemini displays source links to help users verify information and that the company continually tests and improves its models.
For businesses and platforms alike, the challenge will be closing the gap between marketing AI as a shopping assistant and ensuring it doesn't funnel customers toward scams. As adoption grows, the stakes for getting recommendations right will only intensify.