AI chatbots are recommending fake retail websites, study finds

A study by IP protection firm Podqi, shared with The Verge, asked ChatGPT, Gemini and Perplexity shopping questions about 151 brands across 6,795 answers. All three surfaced risky sites: fake stores impersonating retailers, knockoff sellers, reregistered old brand domains or dead domains someone could buy.

Perplexity showed risky sites in 15 percent of answers, ChatGPT in 8 percent and Gemini in 4 percent. Smaller brands were hit about twice as often as large ones. With more people expected to shop through AI this holiday season, that is a direct path to stolen card numbers and wasted money.

We want AI shopping help to work. Comparing products and finding where to buy is exactly the kind of tedious task chatbots can make easier. But a shopping assistant has a higher bar than a trivia bot. If it hands someone a fake store, the cost is not an embarrassing answer. It is a payment to a scammer.

The most troubling example in the study is Perplexity naming the wrong page as Stüssy's official site and calling it a place for guaranteed authentic items. That is the failure in miniature: a confident, friendly answer that puts the system's credibility behind a bad link. Fake regional sites were a weak spot too; the engines sometimes listed regional stores for brands that do not have any.

There are caveats worth stating. Podqi wrote its own prompts rather than studying real user conversations, ran its tests in September, and counts mentions that include warnings. Madhappy is one of its clients, and it sells brand-protection services, so it has an interest in the result. Even so, the gap between products is large and the method is clear enough that the AI companies should take it seriously. Google and OpenAI said they are working on it; Perplexity pointed to its accuracy record.

Where this leads if it scales worries us. As chatbots become the first stop for shopping, scammers will optimize for them the way they once optimized for search ads. Smaller brands, which already struggle with impostors, get hurt first. The fix is not to abandon AI shopping. It is verified merchant data, clear warnings and treating official-store claims as something that must be checked, not guessed.

Podqi tested popular fashion and homeware brands including Lululemon, Birkenstock, Stüssy, Madhappy and Le Creuset. ChatGPT warned users about risky sites it surfaced most often, 24 percent of the time. Podqi's CEO advised verifying a site before entering personal or card details.