The dark side of nurse–generative AI collaboration in the workplace: unveiling work alienation and nurse expediency

Background The integration of Generative Artificial Intelligence (GenAI) into nursing practice is expected to enhance efficiency and decision-making. However, this paradigm shift towards “nurse–GenAI collaboration” also introduces underexplored psychological and behavioral risks. While existing literature often celebrates AI’s productivity gains, the potential “dark side”—including work alienation and counterproductive behaviors—remains a critical gap, particularly regarding the specific mediating mechanisms and individual boundary conditions involved. Objective This study aims to examine the potential dark side of nurse–GenAI collaboration by investigating its associations with work alienation and subsequent nurse expediency. Drawing on the Job Demands-Resources (JD-R) model, we propose a moderated mediation model to examine the dual mediating roles of work meaningfulness (cognitive pathway) and emotional exhaustion (affective pathway), and the moderating effect of GenAI anxiety. Methods A cross-sectional survey design was employed, collecting data from 306 registered nurses across various Chinese hospitals who had used GenAI tools in their work. Participants completed validated measures assessing, nurse-GenAI collaboration, work meaningfulness, emotional exhaustion, work alienation, nurse expediency, and GenAI anxiety. Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to analyze the hypothesized relationships. Results The findings supported all proposed hypotheses. Nurse–GenAI collaboration was negatively associated with work meaningfulness (β = -0.299, p < 0.001) and positively associated with emotional exhaustion (β = 0.231, p = 0.008). The indirect associations between collaboration and work alienation were positive and significant through both work meaningfulness (β = 0.120, p < 0.01) and emotional exhaustion (β = 0.071, p < 0.05), and the direct association was non-significant when both mediators were included. Work alienation, in turn, was strongly associated with nurse expediency (β = 0.668, p < 0 0.001). Finally, GenAI anxiety significantly moderated both first-stage associations (interaction terms: β = -0.262, p < 0.001 for meaningfulness; β = 0.166, p = 0 0.035 for emotional exhaustion): the associations of collaboration with meaningfulness and with emotional exhaustion were stronger at higher levels of GenAI anxiety. Conclusions In this cross-sectional sample of Chinese nurses, higher nurse–GenAI collaboration intensity was associated with lower work meaningfulness and higher emotional exhaustion, which in turn were associated with work alienation and expedient behaviors, particularly among nurses with high GenAI anxiety. They suggest that the psychological and behavioral costs of GenAI integration deserve attention alongside its efficiency benefits, and they offer a nuanced extension of the JD-R model to human–AI collaboration in nursing.