Millions of people now start a shopping search by asking an AI chatbot rather than going to Google. That shift in habit has created a new avenue for fraud that is still widely misunderstood.
The scam works like this. A fraudster builds a website that closely copies a known retailer, complete with product photos, pricing, and branding. They then publish dozens of supporting pages, fake reviews, blog posts, and forum entries across the web, all mentioning the fake site alongside the retailer's name. AI tools that generate shopping answers pull from that web content. When enough of it points to the fake site, the AI starts recommending it.
Security researchers call this AI poisoning. The key point is that it requires no technical hacking of the AI itself. The attacker just writes content the AI will find and repeat. As one analysis put it: "the most common attack right now requires zero technical skill."
The problem is made worse by a separate issue: AI chatbots sometimes generate web addresses that do not actually exist, a behaviour researchers call hallucination. One study of over 18,000 visits sent by ChatGPT found that roughly 3 percent of the recommended URLs led to pages that did not exist on the destination website. A scammer who notices the AI consistently suggesting a non-existent URL for a real brand can simply register that address and set up a fake store there.
The shopper's psychology does the rest of the work. When a human types into Google and sees a list of blue links, there is at least some awareness that they need to evaluate what they click. When an AI chatbot gives a confident, conversational answer and offers a link, most people treat it as a verified recommendation. That trust is precisely what these schemes rely on.
The financial backdrop makes this urgent. US consumers reported losing a record $15.9 billion to fraud in 2025, up from $12.5 billion in 2024. Online shopping fraud was among the most commonly reported categories. AI-referred retail traffic grew by nearly 700 percent during the 2025 holiday season compared to the year before, meaning more purchase decisions than ever are now being influenced by chatbot recommendations.
For businesses, there is also a brand damage dimension. Fake stores impersonating your company are not just stealing from your customers; they are destroying trust in your name. A buyer who orders what they believe is your product, receives nothing or a counterfeit, and then cannot get a refund will blame you, not the fraudster.
The practical rules are simple. Never click a product or retailer link directly from a chat window and then enter payment details. Type the retailer's web address into your browser yourself, or search for it directly. Before buying, confirm the domain is exactly right, not an address with extra letters or words like "sale," "official," or a different country code. If a price looks unusually low compared to what you have seen elsewhere, treat it as a signal to stop and verify. These checks add thirty seconds to a purchase and can save you from handing your card details to a criminal.