LONDON – Millions of online shoppers relying on AI to get topbest deals could be walking into digital traps, as a new study reveals that popular AI chatbots, including ChatGPT, Google Gemini and Perplexity, directed users towards fraudulent shopping websites.
A study by US based intellectual property protection firm Podqi raised fresh concerns over reliability of AI-powered shopping recommendations, revealing that even widely trusted chatbots can provide links to suspicious stores, counterfeit sellers and websites falsely claiming to represent established brands.
Researchers looked into 6,795 responses generated by the AI platforms and found that nearly 9% contained references to dangerous portals, exposing weakness in rapidly expanding world of AI-assisted online shopping.
Among three platforms tested, Perplexity recorded highest proportion of suspicious website references, appearing in 15% of its responses. ChatGPT followed with 8%, while Google’s Gemini performed comparatively better, with potentially unsafe websites appearing in 4% of its answers.
ChatGPT stood out for issuing the most safety warnings, cautioning users in 24% of its responses that involved suspicious websites. The wider findings were nevertheless troubling. Across the three platforms, 57% of responses featuring questionable websites actively recommended them as shopping destinations, while 34% mentioned them without a recommendation. Only 9% warned users to stay away.
The investigation went beyond individual chatbot responses, examining 4,437 website domains associated with AI-generated shopping information. Of these, 718 were flagged as risky, including 351 fraudulent websites and another 367 inactive domains. Some suspicious websites allegedly offered counterfeit merchandise, while others involved expired or re-registered domains that could mislead shoppers searching for genuine brand outlets.
The findings highlight how convincing website names and apparently legitimate shopping links can make it difficult for consumers to distinguish authentic retailers from potential scams.
One of alarming discoveries involved searches for international deliveries and country-specific shopping websites. Such queries produced suspicious website references in 16% of responses, the highest rate identified among the shopping categories examined.
Researchers found cases where chatbots presented supposed regional stores for international brands despite those companies having no official shopping websites in the countries concerned.
OpenAI spokesperson Taya Christianson said the company was working to strengthen safeguards designed to prevent fraudulent websites from appearing in shopping recommendations. Google spokesperson Jessica Lews Board acknowledged that Gemini could produce inaccurate information but stressed that the service provides source links to help users independently verify its answers.
Consumers are using different AI tools to compare prices, locate discounted products and identify retailers offering international shipping.
With holiday shopping driving demand for attractive deals, researchers fear that misleading AI recommendations could expose unsuspecting buyers to fake storefronts, counterfeit goods and possible financial fraud.
ChatGPT Privacy Scare as over Users Pictures Leaked by OpenAI AI Agents
