The New Fraud Frontier: AI Hallucinations Are Directing Shoppers to Fake Retailers
AI-generated image · US National Wire
A new study reveals that major AI chatbots are recommending imposters and dead domains to consumers, turning LLM optimization into a critical security priority for brands.
For e-commerce operators, the promise of AI-driven discovery is quickly colliding with a harsh reality: LLM hallucinations aren't just embarrassing—they are becoming a customer acquisition tool for fraudsters. In my view, brands can no longer treat AI search optimization as a mere marketing exercise; it must be integrated into the corporate security layer to protect both revenue and customer trust.
As The Verge first reported, a troubling study conducted by IP protection firm Podqi tested how Gemini, Perplexity, and ChatGPT handle shopping-related queries. The study, which involved 151 brands ranging from large names like Lululemon, Birkenstock, and Le Creuset to smaller labels like Madhappy and Stüssy, found that all three AI engines directed users to "risky" websites. These risks include sites impersonating real retailers, platforms selling knockoffs, dead domains, or old brand domains that have been reregistered by third parties.
According to The Verge, the failure rate varies significantly across the major players:
* **Perplexity** performed the worst, surfacing risky sites in 15 percent of its 6,795 answers. In one instance, Perplexity identified an incorrect page as the official Stüssy website, labeling it the "best starting point for guaranteed authentic items." * **ChatGPT** displayed risky sites in 8 percent of responses, though it was the most proactive in providing warnings, doing so 24 percent of the time. * **Gemini** showed the fewest risky sites, appearing in 4 percent of answers.
From an operator's perspective, the most alarming finding is that smaller brands are disproportionately targeted. Podqi discovered that queries for larger companies resulted in risky site recommendations about half as often as those for smaller brands. Furthermore, the AI engines frequently hallucinated regional websites for brands that do not actually have them, particularly when prompted about international shipping or regional sites.
As we enter the holiday shopping season, the stakes for brand protection are peaking. Podqi CEO and cofounder Trevor West warned that consumers must verify sources before providing credit card or personal information.
The AI providers have responded to these findings with varying degrees of caution. OpenAI spokesperson Taya Christianson stated that the company is improving safeguards to reduce the likelihood of deceptive sites appearing. Google spokesperson Jessica Levis Boord noted that Gemini is designed to display source links to help users verify information. Meanwhile, Perplexity spokesperson Jesse Dwyer defended the tool's track record on research benchmarks, while acknowledging the system is not 100% accurate.