Benchmark: Gemini Most Likely Among AI Chatbots to Mislead Shoppers

A Product.ai benchmark of 220 shopping queries found Google's Gemini 3.1 Pro Preview fabricates product details in 21% of answers and makes costly pricing errors in 56% of responses. Perplexity was the most accurate, with 3% fabrications and 14% pricing errors. Researchers advise verifying prices on official brand sites.

• Gemini fabricated product details in 21% of responses, worst among platforms
• 56% of Gemini answers contained an error that would cost a buyer money
• Perplexity posted 3% product fabrications and 14% pricing errors
• The test covered 220 questions across 9 categories, each repeated 5 times