AI chatbots are recommending fake retail websites, study finds

A new study by IP protection firm Podqi, shared exclusively with The Verge, tested how ChatGPT, Gemini, and Perplexity handle shopping-related questions about 151 brands, ranging from large companies like Lululemon, Birkenstock, and Le Creuset to smaller brands. Across 6,795 answers, all three AI engines cited or displayed risky sites, defined as fake websites impersonating real retailers, knockoff sellers, reregistered old brand domains, or dead domains anyone could buy. Google's Gemini performed best, surfacing risky sites in 4 percent of answers, while ChatGPT was in the middle at 8 percent and Perplexity was worst at 15 percent. ChatGPT offered warnings about surfaced risky sites 24 percent of the time, and smaller brands were nearly twice as likely to see risky sites in answers compared to larger companies. Prompts about regional brand websites were especially likely to trigger risky results, and the chatbots sometimes listed fake regional sites for brands that have none. The issue is a concern for brands and could pose a risk to consumers, particularly during the holiday shopping season when more people may use AI to shop. Experts and companies advise consumers to verify sources before entering personal or payment information. Google, OpenAI, and Perplexity each responded, saying they take information quality seriously and continue improving safeguards, though Perplexity disputed the framing of the findings.

(Source：The Verge)