AI chatbots steer shoppers to fake websites

Study tests three major chatbots

IP protection firm Podqi tested ChatGPT, Gemini, and Perplexity with shopping questions about 151 brands including Lululemon, Birkenstock, Stüssy, Madhappy, and Le Creuset.

Risky sites defined broadly

Podqi defined risky sites as fake websites impersonating real retailers, sites selling knockoffs, old brand domains reregistered by others, or dead sites someone could purchase and reuse.

Perplexity worst performer

Over six thousand seven hundred ninety-five answers, Perplexity showed risky sites in fifteen percent of responses. Gemini showed them in four percent, ChatGPT in eight percent.

ChatGPT offers more warnings

ChatGPT warned users about risky sites twenty-four percent of the time it surfaced them, better than the other two chatbots at alerting shoppers to potential dangers.

Smaller brands hit harder

Queries about larger companies had risky sites in answers about half as often as queries about smaller brands, leaving lesser-known retailers especially vulnerable to impersonation.

Regional queries trigger more fakes

Certain prompts triggered risky sites more often. AI displayed fake regional sites for brands that do not have any, including naming the wrong page as Stüssy's official website.

“the best starting point for guaranteed authentic items”

Verify before you buy

Podqi CEO Trevor West urged consumers to verify sources before entering personal or credit card information, even when results come from a trusted chatbot interface.

“it is always best to verify the source of that before entering personal information or credit card information”

Companies promise improvements

Google said Gemini is AI and can make mistakes, displaying source links to help verify information. OpenAI said it is improving safeguards to reduce deceptive sites in shopping results.