AdvertiseDonate
AboutRSSClevelandDetroitSan AntonioSan DiegoJacksonvilleWashingtonBostonLouisville
November 16, 2026
Herald Horizon · 6100 Oak Tree Boulevard, Suite 200, Independence, OH 44131
Tech · National

Podqi study: AI chatbots link shoppers to fake retail sites in 15 percent of responses

A test of Gemini, ChatGPT, and Perplexity found that Perplexity directed users to risky domains most frequently, while smaller brands faced higher exposure than major corporations.

By Fatou Girard ·

Illustration for: Podqi study: AI chatbots link shoppers to fake retail sites in 15 percent of responses
Illustration generated for this article, not a news photograph.

Artificial intelligence tools are increasingly used by consumers to research products and locate retailers, but a new analysis suggests these assistants often point users toward illegitimate websites. A study conducted by IP protection firm Podqi tested three major AI chatbots, including Google’s Gemini, OpenAI’s ChatGPT, and Perplexity, to determine how frequently they recommended risky URLs in response to shopping queries.

The research team designed specific prompts asking about 151 brands, ranging from large companies like Lululemon and Le Creuset to smaller fashion labels such as Stüssy and Madhappy. The tests ran over two periods in September and generated 6,795 answers. Podqi categorized risky sites as fake storefronts selling knockoffs, expired domains that have been reregistered by new owners, or inactive sites that could be purchased and reused by third parties.

Perplexity exhibited the highest rate of risky recommendations, displaying such sites in 15 percent of its responses. ChatGPT followed with an 8 percent rate, while Google’s Gemini showed the lowest incidence at 4 percent. The study noted that smaller companies were particularly vulnerable, with queries regarding major brands producing risky links in only half as many instances as those involving smaller firms.

ChatGPT provided warnings about risky sites in 24 percent of cases where they appeared, a higher frequency than its competitors. The data indicated that certain types of questions triggered more errors than others. Specifically, prompts asking for regional brand websites led to a higher number of risky recommendations. In one instance, Perplexity identified the wrong page as Stüssy’s official site, describing it as the best starting point for authentic items even though the brand did not have that specific regional page.

Trevor West, CEO and cofounder of Podqi, advised consumers to verify sources before entering personal or payment details. He stated that users should always check the origin of a link, even when it is provided by a trusted chatbot. This caution is particularly relevant as the holiday shopping season approaches, a time when consumers often seek out new retailers and in-stock items.

Representatives from the AI companies responded to the findings. Jessica Levis Boord, a Google spokesperson, said the company takes information quality seriously and displays source links to help users verify information. Taya Christianson from OpenAI stated that the company is improving safeguards to reduce the likelihood of deceptive sites appearing in shopping results. Jesse Dwyer, a spokesperson for Perplexity, noted that the company does not claim 100 percent accuracy but remains focused on maintaining high standards of reliability.

The study highlights the ongoing challenge for AI developers to balance helpful product discovery with accuracy. As more shoppers rely on these tools for price comparisons and availability checks, the frequency of incorrect or unsafe recommendations becomes a measurable factor in the user experience.

More Tech · Report a business · Got a Story?

Get Herald Horizon in your inboxToday’s news, without the noise. Free, a few times a week, unsubscribe any time.

Get Herald Horizon in your inboxToday’s news, without the noise. Free, a few times a week, unsubscribe any time.