<p>As artificial intelligence (AI) increasingly shapes educational environments, its psychological impact on learners remains underexplored—particularly in relation to emotional well-being. This study explored the longitudinal emotional trajectories of English as a foreign language (EFL) undergraduate students engaged in AI-supported English language learning, focusing on anxiety, depression, and emotional well-being. A total of 734 EFL students from multiple Chinese provinces and diverse academic disciplines completed an online survey. Using latent growth curve modeling (LGCM) within the structural equation modeling (SEM) framework, the analysis revealed significant variances in initial emotional states—anxiety (σ² = 0.36, SE = 0.41, <i>p</i> &lt; 0.05), depression (σ² = 0.52, SE = 0.39, <i>p</i> &lt; 0.05), and emotional well-being (σ² = 0.49, SE = 0.42, <i>p</i> &lt; 0.05)—indicating heterogeneity in baseline emotional conditions. Significant slope variances for anxiety (σ² = 0.45), depression (σ² = 0.60), and well-being (σ² = 0.51) (all <i>p</i> &lt; 0.05) suggested diverse emotional developments over time. Growth pattern correlations revealed that increases in anxiety and depression were associated with declines in emotional well-being. An extended model incorporating AI-driven learning environments showed a good overall fit (χ² = 2.353, df = 31, <i>p</i> = 0.02, CFI = 0.964, RMSEA = 0.071) and identified AI engagement as a significant predictor of emotional change—reducing anxiety (β = −&#xa0;0.51, <i>p</i> &lt; 0.05) and depression (β = −&#xa0;0.46, <i>p</i> &lt; 0.05) while enhancing well-being (β = 0.63, <i>p</i> &lt; 0.05). AI tools accounted for 29% of the variance in anxiety, 21% in depression, and 42% in emotional well-being. These findings underscore the potential of AI learning environments to foster positive emotional development in language learners.</p>