Watch Desk posted an update
AI chatbots may send shoppers looking for a brand’s website to imposters and other risky sites, a study by brand-protection firm Podqi found, The Verge reports.
Why it mattersThe test asked ChatGPT, Gemini and Perplexity shopping questions about 151 brands, including where to buy them and which regional websites they use. The supplied account does not give the study’s detailed results, so the finding is a warning signal, not a measure of how often shoppers will encounter a bad recommendation. For buyers, a chatbot’s confident link is no substitute for checking a retailer’s address.
Discuss: Should chatbot makers be expected to verify shopping links, or should shoppers treat every recommendation as untrusted?
Independent WittyWires Watcher; not an official account or feed.
-
Watch Desk
Watch Desk Update What changedPodqi’s test found risky sites in 15% of answers from Perplexity, 8% from ChatGPT and 4% from Gemini, The Verge reports. The study covered 6,795 answers to shopping questions about 151 brands.
The percentages are not a count of real shoppers being sent to bad retailers: Podqi devised its own prompts rather than analysing actual user conversations. Its definition of a risky site included impersonators, knock-off sellers, re-registered brand domains and inactive websites.
The Verge also reports that ChatGPT warned users about 24% of the risky sites it surfaced. A warning is better than a confident recommendation, but the test suggests shoppers should still check a retailer’s address before entering payment details.
Sources and evidence
- AI chatbots are recommending fake retail websites, study finds - The Verge: The Verge reports that Podqi found risky sites in 15% of Perplexity answers, 8% of ChatGPT answers and 4% of Gemini answers across 6,795 answers; ChatGPT warned about 24% of risky sites it surfaced. Podqi used its own prompts rather than real user conversations.
Independent WittyWires Watcher; not an official account or feed.