A study by IP protection firm Podqi, shared exclusively with The Verge, found that ChatGPT, Gemini and Perplexity all cited or showed risky sites when asked where to buy products from 151 brands, including Lululemon, Birkenstock, Stüssy, Le Creuset and Podqi client Madhappy. Podqi counted fake sites impersonating retailers or selling knockoffs, old brand domains reregistered by someone else, and dead sites that others could buy and reuse.
Across 6,795 answers collected over two testing periods in September, Gemini linked to or mentioned a risky site in 4% of answers, ChatGPT in 8% and Perplexity in 15%, according to the study. ChatGPT warned users about the risky sites it surfaced 24% of the time, the best of the three. Queries about larger companies returned risky sites about half as often, and prompts about regional brand websites triggered them most. In one case, Perplexity named the wrong page as Stüssy's official site. Podqi wrote its own prompts rather than using real user conversations.
Google spokesperson Jessica Levis Boord said Gemini displays source links to help people verify information. OpenAI spokesperson Taya Christianson said the company is improving safeguards to make deceptive sites less likely to appear. Perplexity spokesperson Jesse Dwyer said Perplexity is consistently the most accurate AI on credible research benchmarks and never claims to be 100% accurate.
