AI can't take over your holiday shopping quite yet, a new study suggests. (www.businessinsider.com)

A new study by Product.ai reveals that AI-powered shopping assistants are not yet reliable enough for holiday shopping. Testing free and paid versions of popular large language models (LLMs) like ChatGPT, Claude, Gemini, and Perplexity, researchers posed 220 product-related questions and found a staggering 86% of responses contained factual conflicts regarding prices and specifications. The study highlights significant accuracy shortcomings across all platforms, with Gemini showing the highest rate of costly errors—56% in its free tier, while Perplexity had the lowest at 14%. This research is significant for the AI/ML community as it underscores the current limitations of AI in providing accurate and consistent shopping assistance. Users are cautioned against relying entirely on AI agents, as the discrepancies in pricing could lead to substantial financial mistakes, with some erroneous price quotes off by a median of $300. Product.ai advises that consumers should utilize AI tools as aids for discovery rather than definitive sources, emphasizing the need for verification through multiple sources or direct checks on retailer websites before making purchases.