Study finds AI chatbots sometimes link shoppers to fake stores

In short: A new test found that popular AI chatbots can point shoppers to fake or risky shopping websites.

What's going on

A study by IP protection firm Podqi tested three AI chatbots, ChatGPT, Google’s Gemini, and Perplexity, on shopping related questions for 151 brands. The prompts included questions like where to buy a brand online and whether a brand has regional websites.

Across 6,795 answers, Podqi found the chatbots sometimes showed “risky” website links. Podqi defined risky links as fake websites pretending to be real retailers, sites selling knockoffs, old brand web addresses that someone else re-registered, or dead websites that are no longer active. Think of it like asking for directions and being sent to a storefront that looks real, but is actually a copy.

Perplexity showed risky sites in 15 percent of answers in the test, the highest of the three. ChatGPT showed risky sites in 8 percent of answers. Gemini showed risky sites in 4 percent of answers, the lowest rate in the study.

Podqi also said smaller brands seemed more exposed. Queries about bigger companies had risky sites about half as often. Some prompts appeared to trigger more bad links, especially questions about regional websites.

Google and OpenAI both said they are working to improve results and that users should verify sources. Perplexity said it focuses heavily on accuracy, but does not claim to be perfect.

What to watch

This matters more during holiday shopping, when people may rush to unfamiliar websites for deals or hard to find items. If you use a chatbot for shopping, double check the official brand site before entering a password or credit card number.

Source: The Verge AI