Study Finds AI Chatbots Recommend Risky Retail Websites
A recent study by Podqi revealed that AI chatbots, including ChatGPT, Gemini, and Perplexity, frequently direct users to potentially fraudulent retail websites. The research indicated that up to 15% of chatbot responses included links to risky sites, particularly when users inquired about smaller brands. This issue is especially concerning as the holiday shopping season approaches, increasing the likelihood of consumers relying on AI for purchasing decisions.
1 min read1 source
Key takeaways
30-second read- 1Developers and product teams need to enhance AI training datasets to minimize the risk of recommending fraudulent websites, especially during peak shopping seasons.
- 2Operators should implement better verification processes for AI-generated content to ensure consumer safety and trust.
- 3Retailers must be aware of the potential reputational damage from AI misguidance and consider proactive measures to protect their brands.
Confirmed
No official source linked yet
Reported
Secondary reporting: The Verge AI
Verify yourself
Eligibility, local prices and dates for your car, state or utility

A recent study conducted by Podqi and reported by The Verge highlights a significant issue with AI chatbots, revealing that they often recommend risky retail websites to users. The study tested three popular chatbots—ChatGPT, Gemini, and Perplexity—by posing shopping-related questions about 151 brands. The results were concerning: up to 15% of the chatbot responses included links to potentially fraudulent sites, which could mislead consumers during their online shopping experiences.
The study found that Perplexity had the highest rate of risky recommendations at 15%, while ChatGPT and Gemini followed with 8% and 4%, respectively. Notably, ChatGPT was recognized for providing warnings about risky sites 24% of the time, which could help mitigate some risks for users. The findings indicate that smaller brands are particularly vulnerable, with queries about larger companies yielding risky site recommendations only about half as often.
As the holiday shopping season approaches, the implications of these findings are significant. Developers and product teams must prioritize improving AI training datasets to reduce the likelihood of chatbots directing users to fraudulent websites. This is crucial as more consumers are expected to rely on AI for shopping, potentially leading to increased instances of identity theft or financial loss if they engage with these deceptive sites.
Operators of AI systems should also consider implementing stronger verification processes for the information generated by their models. This could involve cross-referencing AI outputs with reliable databases or enhancing the algorithms to better discern legitimate sources from fraudulent ones. Retailers, on the other hand, need to be proactive in safeguarding their brands against the reputational damage that could arise from AI misguidance.
Podqi's CEO, Trevor West, emphasized the importance of consumer awareness, advising users to verify the sources of information before providing personal or financial details. As AI continues to evolve and become more integrated into shopping experiences, ensuring the accuracy and reliability of its recommendations will be paramount.
Sources
1 media- [1]AI chatbots are recommending fake retail websites, study findsMedia
The Verge AI · theverge.com · published Nov 12
Newsletter
Weekly AI signal. No spam.




Comments
Sign in to join the discussion
Loading comments…