This quasi-experimental mixed-methods study investigated the effects of AI-assisted reading instruction on foreign language reading anxiety and ambiguity tolerance among university-level EFL learners. Drawing on Cognitive Load Theory, Attentional Control Theory, and Positive Psychology, the study examined whether structured use of GPT-4–based text simplification and feedback mechanisms could alleviate affective barriers while enhancing cognitive flexibility in reading. A total of 120 intermediate EFL students were assigned to experimental (n = 60) and control (n = 60) groups over a 12-week intervention. The experimental group engaged in AI-supported reading sessions involving adaptive text simplification, contextual clarification, and immediate formative feedback, while the control group received traditional instruction. Quantitative data were collected using the Foreign Language Reading Anxiety Scale (FLRAS) and the Second Language Ambiguity Tolerance Scale (SLATS), complemented by semi-structured interviews. Results revealed a statistically significant reduction in reading anxiety and a substantial increase in ambiguity tolerance in the experimental group, with no comparable gains in the control group. Correlational analyses further indicated that reductions in anxiety were associated with improvements in ambiguity tolerance and reading performance. Qualitative findings corroborated these results, highlighting enhanced reading confidence, perceived emotional safety, and greater resilience in dealing with textual uncertainty. The findings suggest that AI-supported reading environments may simultaneously address cognitive load and affective constraints, thereby fostering more adaptive and confident EFL readers. Implications for AI integration in language pedagogy and future longitudinal research are discussed.