Enhancing English Reading Comprehension, Learning Motivation and Attitude Through AI ‐Supported Pre‐Reading Scaffolding

ABSTRACT Background The integration of generative artificial intelligence (Gen‐AI) into second language (L2) education has opened new possibilities for personalized and adaptive instruction. While Gen‐AI has shown promise in supporting language production skills, its application in reading, particularly in pre‐reading scaffolding, remains underexplored. Objectives This study aimed to examine the effects of AI‐supported pre‐reading scaffolding, using ChatGPT, on university EFL students' English reading comprehension, learning motivation, and attitudes towards English learning, compared to traditional teacher‐led scaffolding. Methods A quasi‐experimental pretest‐posttest design was employed with 66 first‐year university EFL students. Participants were assigned to either a teacher‐led control group ( n = 34) or an AI‐supported experimental group ( n = 32). Both groups received structured question‐posing scaffolding prior to reading tasks, but delivery methods differed. The control group interacted with instructors, while the experimental group used a custom‐designed online platform integrating ChatGPT. ANCOVA was used to analyse the post‐intervention outcomes across three domains: reading comprehension, motivation, and attitudes. Results ANCOVA results indicated that students in the AI‐supported group outperformed the teacher‐led group in English reading comprehension, F (1, 63) = 5.06, p = 0.02, partial η 2 = 0.07. Similarly, learning motivation was higher in the AI group, F (1, 63) = 5.90, p = 0.01, partial η 2 = 0.08. For learning attitude, the AI‐supported group also demonstrated an advantage, F (1, 63) = 14.77, p < 0.001, partial η 2 = 0.19. These results suggest a moderate‐to‐large effect of AI‐supported pre‐reading scaffolding on both cognitive and affective outcomes. Conclusions Integrating Gen‐AI into pre‐reading scaffolding enhances both cognitive and affective outcomes in EFL reading contexts. These findings support the potential of AI‐based instructional tools to provide scalable, interactive, and learner‐centered support for reading development in higher education.