Latent profiles of AI teaching self-efficacy and their associations with occupational well-being among vocational education teachers in China: the moderating role of AI anxiety

The rapid diffusion of artificial intelligence (AI) tools has placed Chinese vocational education teachers at the frontline of a transition combining instructional, practical-training, and emotional demands. Although prior variable-centered research has linked teachers’ AI-related self-efficacy to occupational outcomes, heterogeneity in the multidimensional structure of AI teaching self-efficacy has not been systematically examined in vocational education, and the moderating role of AI anxiety in linking such configurations to occupational well-being remains untested. A regionally stratified online survey was conducted with 768 in-service teachers (378 from secondary vocational schools, 390 from higher vocational colleges) across eastern, central, and western mainland China. AI teaching self-efficacy was measured with the six-dimensional Teacher Artificial Intelligence Competence Self-Efficacy Scale (TAICS), AI anxiety with a teaching-contextualized adaptation of the AI Anxiety Scale, and occupational well-being with the Teacher Subjective Well-Being Questionnaire (TSWQ). Latent profile analysis was conducted in Mplus alongside multi-group confirmatory factor analysis, covariate (R3STEP) analysis, and posterior-probability-weighted (BCH) distal-outcome analysis. Full scalar measurement invariance of the TAICS held across school levels. Four distinct profiles emerged: Foundational Stage (21.4%), Technically Skilled but Ethically Cautious (21.4%), Pedagogically Engaged but Technically Constrained (25.7%), and Holistic Competence (31.6%). Occupational well-being rose monotonically across profiles (TSWQ M = 2.74, 3.04, 3.17, 3.43), with pairwise Cohen’s d (pooled within-class SD) from 0.39 to 2.06. AI anxiety moderated the profile-to-well-being association only for the Holistic Competence profile (B = − 0.13, p < .001); the interaction did not reach significance for the Technically Skilled (B = − 0.07, p = .062) or Pedagogically Engaged profiles, and it accounted for a small share of variance overall (ΔR² = 0.012). The Holistic-over-Foundational well-being advantage was correspondingly smaller, by roughly one third, across the observed range of AI anxiety, an association we interpret cautiously given the cross-sectional design. Chinese vocational education teachers organize into meaningful configurations of AI teaching self-efficacy, with broadly balanced configurations conferring the largest well-being benefits but showing the steepest anxiety-related attenuation of that advantage. The findings apply the Job Demands–Resources framework to AI-in-education research and support differentiated development and support strategies matching competence-building and anxiety-mitigation interventions to teachers’ profiles.