Shopping chatbots sometimes point people towards websites that impersonate retailers or sell knock-offs, according to findings from a study by Podqi reported by Futurism. In the results described, Perplexity produced risky sites in 15% of its answers, ChatGPT in 8% and Gemini in 4%. The figures are a useful warning: a polished answer is not the same thing as a trustworthy shop.
Watch Desk analysis
What happened
Futurism says the study tested responses to questions about where to buy branded products, whether brands ship internationally and which regional websites they use. It covered about 151 brands, from smaller names to well-known companies including Lululemon and Birkenstock. Futurism reports the study was conducted by Podqi and reported on by The Verge.
The reported results distinguish showing a risky site from endorsing it. Across answers that surfaced one, 57% recommended it as a place to shop, while 34% did so without comment and 9% warned users away. ChatGPT included warnings in 24% of its responses that showed risky sites, the highest share among the three examples given.
Why it matters
A chatbot can serve up a dubious storefront even when the question is ordinary shopping research. The risk may be greater for smaller brands, which the study reportedly found were more likely to be displaced by scam sites. A shopper who treats an answer as a verified recommendation could end up sharing payment or personal details with the wrong seller.
Our read
The numbers are a reason to check the destination, not a league table proving one chatbot is safe. The results are reported through Futurism’s account of a Podqi study, and the reported rates describe this test, not every shopping query. Still, the practical lesson is pleasantly unglamorous: before entering card details, check that the site is the retailer’s real one. AI can make finding a shop quicker; it has not made due diligence obsolete.
What to watch
- Whether Podqi or The Verge publishes the full methodology and results.
- Whether the chatbots change how they label or filter retail recommendations.
- Whether follow-up testing finds similar rates across different brands and query types.
Discussion spark: When a chatbot names a shop, should it be responsible for checking that the site is legitimate, or is that still the shopper’s job?
Sources and evidence
- Source update (14 November 2026, 18:02 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.