The rapid diffusion of artificial intelligence (AI) is profoundly reshaping the labor market while triggering deep-seated “replacement anxiety” among workers. Exploring the mechanisms underlying this anxiety is of great significance for social risk governance in the digital era. Based on data from the Chinese Social Survey (CSS), this study combines ordinary least squares (OLS) regression with machine learning models to examine the determinants of AI replacement anxiety and its underlying mechanisms. The findings reveal the following: First, the overall level of AI replacement anxiety among Chinese workers is at a low-to-moderate level but exhibits significant structural differentiation, with platform workers and non-state-sector workers reporting higher levels of anxiety. Second, OLS regression identifies a general protective mechanism, demonstrating that state-sector employment and higher socioeconomic status significantly mitigate individuals’ replacement anxiety. Third, machine learning predictions focused on the high-anxiety cohort further indicate that personal income and years of education are core features for identifying highly anxious individuals, suggesting that a lack of basic economic resources and human capital is a key trigger for severe technological anxiety. Interaction analysis further shows that hukou status and education exert a synergistic effect, while income and class status have an antagonistic effect, jointly shaping the formation of replacement anxiety. This study deepens the empirical understanding of the unequal distribution of technological risks in the digital age, providing evidence for alleviating public employment anxiety and formulating inclusive public policies.