AI chatbots are recommending fake retail websites, study finds

What happened

Podqi, an intellectual property protection firm, conducted a study in September testing ChatGPT, Gemini, and Perplexity chatbots with shopping-related questions about 151 brands. The study found that all three chatbots sometimes recommended or displayed links to fake or risky retail websites, including sites impersonating real retailers or selling counterfeit goods. Perplexity showed risky sites most often, followed by ChatGPT and then Gemini. The issue was more pronounced when users asked about smaller brands or regional brand websites.

Key details

• Podqi tested 6,795 chatbot answers across 151 brands.
• Perplexity recommended risky sites in 15% of responses, ChatGPT in 8%, and Gemini in 4%.
• ChatGPT gave warnings about risky sites 24% of the time when they appeared.
• Smaller brands were more likely to have risky sites recommended than larger brands.
• Prompts about regional brand websites triggered more risky site recommendations.

Why it matters

Consumers using these chatbots for shopping advice may be directed to fraudulent or inactive websites, increasing the risk of scams or purchasing counterfeit goods. Smaller brands are especially vulnerable, as they are more likely to have fake sites recommended in chatbot responses. Retailers and shoppers now face the need to verify site authenticity before making purchases or sharing personal information.