AI chatbots are directing shoppers toward fraudulent or inactive websites, according to a study by IP protection firm Podqi. Testing conducted in September across 6,795 answers found that ChatGPT, Gemini, and Perplexity all cited "risky" URLs in various shopping-related queries involving 151 brands, including Lululemon, Birkenstock, Stüssy, and Le Creuset.
Podqi defined risky sites as those impersonating real retailers, selling knockoffs, or using old brand domains that have been reregistered by third parties. The study found that Perplexity performed the worst, displaying risky sites in 15 percent of answers. ChatGPT surfaced risky links in 8 percent of cases, while Google’s Gemini showed them least often at a rate of 4 percent. The research noted that queries regarding regional brand websites were more likely to trigger these results, sometimes resulting in the AI naming fake regional sites for brands that do not actually have them.
Smaller brands were found to be at higher risk than larger companies, with queries for major brands receiving risky site recommendations about half as often as those for smaller entities. While ChatGPT provided warnings about risky sites in 24 percent of the instances where it surfaced them, other models did not offer similar cautions.
"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.
Google spokesperson Jessica Levis Boord stated that the company works to improve its models and noted that Gemini displays links to sources to help users verify information. OpenAI spokesperson Taya Christianson said the company is working to improve safeguards to make deceptive sites less likely to appear in shopping results. Podqi spokesperson Jesse Dwyer added that the company is focused on information quality and accuracy.
