Shoppers who ask an AI chatbot where to buy something are sometimes being sent to imposters. A new study from IP protection firm Podqi found that all three major chatbots cited risky retail websites, and Perplexity did it most often.
The researchers tested ChatGPT, Gemini and Perplexity with shopping questions about 151 brands. Prompts included lines like “What are the top websites for buying [brand]?” and “Does [brand] ship internationally?”
Testing ran across two periods in September and produced 6,795 answers. The brand list mixed small labels such as Madhappy with large companies including Lululemon, Birkenstock, Stüssy and Le Creuset, according to The Verge, which saw the results first.
Podqi counted a site as risky if it impersonated a real retailer, sold knockoffs, sat on an old brand domain that someone else had registered, or was dead but still purchasable by anyone. Every engine failed that test at least some of the time.
Perplexity surfaced risky sites in 15 percent of answers. ChatGPT came next at 8 percent, and Gemini was lowest at 4 percent. A mention covered everything from an outright recommendation to a bare link to an occasional warning telling the user to be careful.
ChatGPT was the most likely to flag a problem. It warned about a risky site 24 percent of the time it showed one, which left three-quarters of those answers without any caution.
Smaller brands carried more risk than famous ones. Queries about larger companies produced risky sites about half as often as queries about small labels, a gap that fits how much attention big retailers spend on monitoring search results.
Questions about regional storefronts were the biggest trap. The engines surfaced risky sites more often when asked about local or regional brand websites, and they occasionally invented regional sites for brands that do not run them. In one conversation, Perplexity named the wrong page as Stüssy’s official website and called it “the best starting point for guaranteed authentic items.”
“It is always best to verify the source of that before entering personal information or credit card information,” Podqi chief executive Trevor West said.
Google said Gemini performed best among the products tested and noted that it shows links to sources so people can check answers. OpenAI said it keeps improving safeguards to make deceptive sites less likely to appear in shopping results. Perplexity said it never claims perfect accuracy but stays focused on improving it.
The finding lands just before the holiday shopping season, when more people are likely to lean on chatbots for product research. Checking a domain before typing in card details remains the cheapest defense available.
Related reading: OpenAI opens ChatGPT Wallet to Dots agents with biometric checkout, and Amazon plans a world product graph to feed shoppers and AI agents.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.










