Large language models (LLMs) are transforming higher education, yet their learning benefits likely depend on students’ AI literacy and how they prompt and verify AI outputs. This mixed-methods study investigates how AI literacy relates to prompting proficiency and verification behavior, and how these behaviors operate through two mechanisms, trust calibration and extraneous cognitive load, to predict coursework performance across two discipline–institution contexts in Thailand. The author surveyed Design students (n = 221) and Business students (n = 222) and analyzed the data using multigroup structural equation modeling, complemented by thematic analysis of semi-structured interviews. Results indicated that AI literacy significantly predicted both prompting proficiency and verification behavior (p < .001). Prompting proficiency was positively associated with trust calibration, whereas verification behavior was negatively associated with extraneous cognitive load, and the corresponding indirect effects were supported. Context-specific patterns also emerged: in the Design context, verification behavior showed a stronger direct association with task quality, while in the Business context, trust calibration was the stronger predictor of assignment quality. Interview evidence helped explain these differences. Design students used ChatGPT primarily to accelerate ideation and refinement, whereas Business students used it to support structured analysis and to make their work more auditable. Across both contexts, verification functioned as a metacognitive safeguard that reduced overload and stabilized accuracy. These findings suggest that teaching AI literacy in higher education should emphasize rubric-guided prompting, verification routines, and calibrated trust to balance efficiency with deeper engagement, with attention to context-specific pedagogical demands.
- Inicie sesión para enviar comentarios