AI-generated self-assessment as a metacognitive scaffold in EFL learning

Abstract

The present study investigates the relationship between English as a Foreign Language (EFL) undergraduates' use of AI-generated self-assessment (AIGSA) and their reported metacognitive skill development. A survey was administered to examine the perceptions of 242 EFL students in Saudi Arabia and Yemen regarding AIGSA and to assess their perceptions of the validity of AIGSA in comparison with instructors' assessments. The quantitative results indicated that EFL students have a high perception of the use of AIGSA in their learning (M = 3.97, SD = .622). The study also revealed a moderate positive correlation between the perceived use of AIGSA and the reported development of EFL students' metacognitive skills (r = .432). The analysis revealed that nationality exerted a significant role in students' perception, where Saudi EFL students reported a higher perception (M = 4.05) than their Yemeni counterparts (M = 3.89, Z = 2.166, p = .001). Moreover, the thematic analysis indicated that AIGSA supports learners' self-monitoring and regulation through accessible, detailed, and low-pressure feedback, while its perceived validity remains contextual and complementary to teacher assessment. By foregrounding learners' evaluative knowledge and trust in feedback sources, this study offers novel insights into how AIGSA functions as a metacognitive scaffold rather than a substitute for teacher feedback.