Abstract:

Generative AI is increasingly used by consumers to search and compare products online. Although previous research has examined trust in AI recommendations and the use of large language models in shopping tasks, less attention has been given to how AI systems compare different types of product information when no product is complete on all dimensions. This study explores whether the information prioritized by an AI shopping assistant changes according to the shopper’s stated purpose. Three merchant-controlled information dimensions were examined: Product Information Completeness, Trust Evidence, and Purchase Readiness. A controlled experiment was conducted using GPT and Gemini under four shopper-intent conditions: neutral, authenticity concern, immediate purchase, and specification-focused. Forty independent recommendation sessions were completed. The results indicate that shopper intent was reflected in the final recommendations of both systems. The clearest pattern appeared when the customer intended to purchase immediately, where alternatives containing price, availability, and delivery information were consistently preferred. The study provides preliminary evidence that AI shopping recommendations are influenced not only by the available product information but also by the way the shopper expresses the decision goal.