Abstract: This study compares the differences between generated AI supported conversational search and traditional web search in user information search behavior and learning outcomes. Combined with the cognitive load theory, this study explores the information acquisition efficiency, learning outcomes and user experience of users using the two systems in various learning search situations, aiming to deeply understand the characteristics of user information search behavior and help improve the information search and learning outcomes. The experimental method was used to collect the index data, the experimental subjects were divided into a conversational search and a traditional web search group, and the user's learning information search behavior was divided into information seek behavior, information selection behavior and information utilization behavior. The learning search outcomes were divided into five dimensions, and a linear regression model was used to compare the differences in information search behavior and learning effect between the two groups of users. Results show that users' use of conversational search system can improve the learning outcome by reducing the cognitive load, and also make part of the user's information search behavior more complex. This study provides theoretical support for revealing the cognitive mechanism behind users' information search behavior and learning outcome differences, broadens the research vision of "search as learning", and provides practical enlightenment for improving the interactive design of the functions of conservational search and traditional web search system and providing learning information service functions.

Key words: Conversational search, Search as learning, Information search behavior, Learning outcome；Comparative analysis