Earlier this year, a retailer called Russell & Bromley stopped operating as an independent business. It happened after they went into financial collapse and got absorbed by Next. However, people have been searching for the business online lately, and scammers took notice.
They then created websites designed to look like the old retailer's official site. Researchers found cases where ChatGPT's shopping-related responses pointed users toward some of these fraudulent stores. Here's what happened in full.
ChatGPT got a little too confident with its recommendations
There are many places to get shopping recommendations, these days. A chatbot should not be among them. ChatGPT was found recommending links to fake online stores in some shopping-related queries, proving that AI should be treated as a starting point for research rather than a crutch for making important decisions.
According to findings cited by The Guardian and scam-detection service Ask Silver, some users searching for products or specific retailers through ChatGPT were presented with links to the websites that appeared genuine.
A user asked the tool what the popular Russell & Bromley bags were, and it responded by listing categories of oversized shoulder bags, hobos, camera bags, and crossbodies. Among them was an “Everyday Oversized Shoulder Bag” with a £275 ($369) price tag.
It's worth knowing that Russell & Bromley had ceased operating as an independent retailer after going bust in January 2026 and subsequently being absorbed by Next. But continuous search from consumers may have continued searching for the brand's former website drew attention to the business and got scammers making copycat pages to fill the gap.
In short, AI poisoning.
As large language models and other AI systems draw on information available across the web, it's not out of the question that they'll also draw in fake recommendations. Some of the flagged websites were reportedly removed from ChatGPT's shopping results after being brought to OpenAI's attention.
Cases like this are not entirely new. Years before the recent fake-store controversy, researchers and students were already discovering that the chatbot could confidently invent sources that never existed.
In one widely shared Reddit post, a PhD student described how the bot cited a supposedly relevant 2019 research paper. It provided a convincing summary of its findings, and even generated a link. However, the paper did not exist.
ChatGPT is not your best friend
Although ChatGPT has removed the fake websites, the underlying problem remains. Fraud experts advise users to verify retailer websites independently, check web addresses carefully, be wary of unusually large discounts, review contact information, and avoid assuming that any website is trustworthy without these scrutiny.
If you already entered payment card details on any suspicious websites, contact your bank or card issuer immediately and ask them to freeze the card or block transactions. They may cancel it and issue a new one. If any transactions have already gone through, report them as unauthorized so they can be reversed or investigated.
If you also entered login credentials, change your passwords, especially if you reuse them across different accounts. Next, turn on two-factor authentication to block further access even if the password has been exposed.
