Artificial intelligence (AI) technologies, particularly AI-Generated Summaries (AIGS), are profoundly reshaping the user interaction ecosystem and decision-making processes on content service platforms by altering the way information is presented and processed. Grounded in the Cognitive Miser Theory, this study employed a field experiment on a leading Chinese video social media platform to examine the impact of AIGS on user endorsement behaviors (e.g., likes，coin-tipping, and favorites) and commenting behaviors. We find that AIGS reduces users’ cognitive load in watching and comprehending videos, thereby positively influencing their endorsement and commenting behaviors. Further analysis reveals that AIGS also significantly alters the hierarchical structure of video comments. From the cognitive perspective, this study examines how AIGS, as an information processing tool, can induce potential double-edged sword effects, thereby enriching the existing theoretical research on AI applications and platform governance.