<p>What makes students prefer to self-disclosure to AI teaching assistant (AI TA) than human teaching assistants? Understanding this question can help online learning platforms better understand students’ needs in the data collection stage to provide more appropriate recommendations and deploy AI and human-AI collaboration more effectively. Based on stimulus-organism-response (SOR) theory, using semi-structured interviews and a quantitative survey, this study explores the influencing factors and underlying mechanisms of students’ preferences for self-disclosure to AI TA. The results show that AI TA has better perceived privacy, fairness, and service attitude than human TA, so students have higher psychological safety when interacting with AI TA and prefer self-disclosure to AI TA. This study also explores the effects of relationship norms on self-disclosure and finds that students are more willing to self-disclosure to AI TA when they want to establish communal relationships with it. These results provide suggestions for the deployment of AI TA and the design of human-AI collaboration and provide new enlightenment for self-disclosure, AI TA, and human-AI interaction research.</p>