23 Daily — According to theverge.com, a recent study by intellectual property protection firm Podqi revealed that shoppers using AI chatbots such as ChatGPT, Google Gemini, and Perplexity to research products or find buying options may be directed to counterfeit or risky websites, posing consumer safety concerns. The study conducted in September 2026 tested responses from these three AI chatbots with questions related to 151 brands, covering topics such as top purchasing websites, finding brand products online, and international shipping details.
The brands tested ranged from fashion names like Lululemon, Birkenstock, Stüssy, and Madhappy (a Podqi client) to homeware brand Le Creuset. Podqi crafted its own prompts, rather than using real user conversations, and collected 6,795 answers over two testing periods. The firm found that all three AI engines cited or linked to risky sites, defined as fake retailers, knockoff sellers, expired brand domains now re-registered by others, or inactive sites that could be repurposed maliciously.
Frequency and differences in AI recommending risky or counterfeit websites
Among the AI chatbots tested, Google Gemini showed risky sites least frequently at 4% of answers, including recommendations, displaying without comment, and user warnings. Perplexity performed the worst, linking to risky sites in 15% of responses, while ChatGPT was in the middle at 8%. ChatGPT also had the highest rate of issuing warnings when mentioning risky sites, about 24%.
Smaller brands faced higher risks, with Podqi reporting that answers related to larger companies contained risky sites about half as often. Queries about regional brand websites triggered more instances of fake regional sites being suggested. For example, Perplexity incorrectly identified a non-official page as Stüssy’s official website and called it “the best starting point for guaranteed authentic items.”
Responses from AI companies and ongoing improvements
Google spokesperson Jessica Levis Boord said Gemini is continuously being tested and improved and transparently displays source links alongside responses to help users verify information, acknowledging that AI can make mistakes. OpenAI spokesperson Taya Christianson emphasized ongoing efforts to strengthen ChatGPT safeguards to reduce deceptive site recommendations and expressed appreciation for research highlighting areas of improvement. Perplexity's spokesperson Jesse Dwyer noted their AI ranks highly on credibility benchmarks but does not claim to be 100% accurate, with continuous focus on accuracy.
Editor's take: Potential risks of AI shopping assistants and the need for verification
As AI chatbots become a key tool for consumers researching purchases, their tendency to recommend counterfeit or risky websites exposes a security blind spot in ecommerce. The likelihood of risky site exposure varies by brand size and regional specificity. With holiday shopping seasons approaching and increased consumer reliance on AI to find deals and stock, risks rise accordingly.
Entrepreneurs should note opportunities to develop more trustworthy AI shopping assistants featuring counterfeit detection and user-friendly reputation verification capabilities, enhancing consumer protection. Practical verification includes checking the authenticity of AI-suggested purchase links by domain validation, comparing with official brand websites, and consulting third-party consumer review platforms to safeguard personal and payment information.