Cognitive gains and addictive losses: The double-edged sword of Generative AI on employee creative thinking

Drawing on Social Cognitive Theory (SCT), this study examines how Generative Artificial Intelligence (GAI) use is associated with employee creative thinking through creative self-efficacy and GAI addiction, and whether GAI use satisfaction moderates this process. A cross-sectional questionnaire survey was conducted among 194 knowledge workers in China. Confirmatory factor analysis was performed using AMOS 29.0, and the mediation, serial mediation, and moderation effects were tested using SPSS 26.0 and the PROCESS 4.2 macro with 5000 bootstrap samples. The results show that GAI use is positively associated with employee creative thinking, and creative self-efficacy mediates this relationship. Meanwhile, the tolerance dimension of GAI addiction is negatively associated with creative thinking, whereas the withdrawal pathway is not significant. Creative self-efficacy also negatively predicts both withdrawal and tolerance, suggesting its protective role in reducing GAI addiction risk. In addition, GAI use satisfaction strengthens the positive relationship between GAI use and creative self-efficacy. These findings indicate that the relationship between GAI use and creative thinking is not uniformly positive or negative, but depends on employees' ability beliefs, addiction-related risks, and affective evaluations of human-AI collaboration. This study provides an integrated perspective on the cognitive empowerment pathway, dependence risk pathway, and affective boundary condition of GAI use in workplace creativity.