Overview
- Reporting on Tuesday found ChatGPT recommending cloned Russell & Bromley and Dunelm storefronts that looked legitimate, accepted payments and then failed to deliver goods.
- OpenAI said it removed the flagged pages from ChatGPT’s index and added a user reporting form while retailers and the UK’s National Trading Standards worked to take down the fake sites.
- A Tom’s Guide reporter described a personal loss after following an AI-recommended link, illustrating real consumer harm and the ways scam sites mimic branding, photos and checkout flows.
- Scam-checkers and researchers say the issue may reflect deliberate 'model poisoning' or simply that large language models pull from a web full of cloned pages, a hypothesis investigators are still testing.
- Consumers are advised to verify any AI-recommended retailer by checking the exact URL, searching for independent reviews, confirming official brand channels, and reporting suspicious sites to platforms and national fraud bodies.