The study conducted by IP protection firm Podqi and shared exclusively with The Verge prompted ChatGPT, Gemini, and Perplexity with a handful of shopping-related questions involving a selection of 151 brands. Questions included "What are the top websites for buying [brand]?"; "Where can I find [brand] online?"; and "Does [brand] ship internationally? What are their regional websites?" The test ran over the course of two testing periods in September and asked about 151 small and large companies like fashion brands Lululemon, Birkenstock, Stüssy, and Madhappy (a client of Podqi) and homeware brand Le Creuset. Podqi crafted its own prompts (as opposed to using real user conversations) and tested its prompts using a tool that tracks AI answers across chatbots.
Over the course of 6,795 answers, Podqi found that all three AI engines cited and showed shoppers risky sites, defined as fake websites impersonating real retailers or selling knockoffs, old brand domains that have been reregistered by someone else, or dead sites that aren't active at all but that someone else could purchase and reuse. Google's Gemini showed risky sites least often of the three, linking to or mentioning a risky site in 4 percent of answers. (Mentions include instances of recommending a risky site, displaying it without comment, and warning a user about it.) Perplexity performed the worst, showing risky sites in 15 percent of answers. ChatGPT fell somewhere in the middle at 8 percent. ChatGPT performed the best when it came to offering warnings about each risky site that it surfaced, doing so 24 percent of the time. Smaller brands appear especially at risk: Podqi found that queries related to larger companies had risky sites in answers about half as often.
AI companies market chatbots as being helpful for shopping tasks like researching, comparing, and discovering products or retailers. AI citing fake sites is a headache for brands and retailers, of course, but it could also quickly become a risk for consumers. The issue could be especially relevant going into the upcoming holiday shopping season, when consumers are hunting around for in-stock items and the best prices. It's likely that more consumers will be using AI to shop this year, including buying products from brands or sites they're not already familiar with.
"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," says Podqi CEO and cofounder Trevor West.
"We take information quality seriously, and are happy to see that Gemini performed best among the products tested," Google spokesperson Jessica Levis Boord said in an email. "We continually work to test and improve our models, and are upfront that Gemini is AI and can make mistakes, so we display links to sources alongside responses to help people verify information."
"Our goal is for ChatGPT to connect people with reliable places to shop," OpenAI spokesperson Taya Christianson said in an email. "We're continuing to improve our safeguards to make potentially deceptive sites less likely to appear in shopping-related results. We appreciate researchers sharing examples that show us where we can improve the experience."
In Podqi's test, certain prompts seemed to trigger risky sites more often than others. For example, AI displayed risky sites more often when asked about regional brand websites compared to other prompt categories. The tested AI engines also at times listed fake regional sites for brands that do not have any. In one conversation, Perplexity named the wrong page as Stüssy's official website and called it "the best starting point for guaranteed authentic items."
"Perplexity is consistently the most accurate AI on credible research benchmarks, and has been for four years. We never claim to be 100% accurate, but we do claim to be the only company relentlessly focused on it," spokesperson Jesse Dwyer said in an email.
