Introduction

Shoppers increasingly turn to artificial intelligence to compare prices, locate retailers, and discover products. But new research suggests these AI helpers may not always point users to legitimate stores. A recent analysis found that popular chatbots sometimes surface fraudulent or inactive retail URLs, creating potential hazards for anyone ready to make a purchase.

What Happened

An intellectual property protection firm called Podqi conducted a controlled test involving three major AI models: OpenAI's ChatGPT, Google's Gemini, and Perplexity. Researchers submitted shopping-related questions about 151 different brands, ranging from fashion labels like Lululemon, Birkenstock, Stüssy, and Madhappy to home goods company Le Creuset. Across two rounds in September, the systems generated nearly 6,800 responses.

The findings showed that all three chatbots referenced what the researchers defined as risky websites. These included counterfeit storefronts impersonating real retailers, lapsed domain names reclaimed by third parties, and inactive pages that could be repurposed by bad actors. Perplexity had the highest rate, appearing in roughly 15 percent of its answers. ChatGPT landed in the middle at about 8 percent, while Google's Gemini showed the lowest frequency at around 4 percent.

The study also noted that prompts asking about regional or international websites triggered risky results more frequently. In one example, Perplexity identified an incorrect page as Stüssy's official site and described it as a reliable source for authentic merchandise.

Why This Matters

AI companies often promote their chatbots as useful tools for product research and retailer discovery. When those tools suggest fake shops, the consequences extend beyond brand reputation. Consumers who follow these links could unknowingly share payment details or personal data with fraudulent operators.

The timing is especially relevant as the holiday shopping season approaches. More shoppers may rely on AI to find hard-to-get items or compare deals, including from retailers they have never used before. Smaller brands appeared to face elevated risk, with risky links appearing about half as often for larger companies.

ChatGPT did offer warnings about risky sites in roughly a quarter of cases, but researchers caution that users should not assume every AI-generated link is safe.

Key Takeaways

  • All three tested AI models - ChatGPT, Gemini, and Perplexity - cited risky or fraudulent retail URLs during the study.
  • Perplexity showed risky sites most often, in about 15 percent of responses, followed by ChatGPT at 8 percent and Gemini at 4 percent.
  • Questions about regional brand websites were more likely to produce unsafe links than other shopping queries.
  • Smaller brands were disproportionately affected compared with major retailers.
  • Experts recommend verifying any website before entering credit card or personal information, regardless of which AI tool recommended it.

Conclusion

The study serves as a reminder that AI shopping assistants are still prone to errors and can be misled by deceptive domains. While the companies involved said they are working on safeguards and improving information quality, the safest approach for consumers remains independent verification. Before completing a purchase through an AI-suggested link, shoppers should confirm the retailer's official domain and look for clear signs of a secure, legitimate checkout process.