[Purpose/significance] The increase in cognitive load during generative information search may lead to user frustration and system-switching behaviour. This paper aims to investigate the patterns and causes of changes in users′ cognitive load at different stages of generative information search. [Method/process] Cognitive load data was collected using experimental methods whilst users were performing search tasks. Cognitive load was measured using subjective, task performance and physiological measurement methods, and a one-way analysis of variance combined with Welch′s test was employed to determine whether there were significant differences in cognitive load across different stages of the search process. Subsequently, subjective descriptions from users were gathered through interviews following the experiment, and thematic analysis was used to identify the causes of cognitive load during the search process.[Result/conclusion] Users′ cognitive load is heaviest during the post-focus stage, followed by the pre-focus stage,with the lightest load occurring during the focus stage. Cognitive load primarily stems from four sources: the user themselves, the task, the generative AI system, and the interaction between the user and the system.