Effectiveness of AI-generated visual scaffolds for reading comprehension in Chinese inclusive classrooms

Reading comprehension remains a significant challenge for students with reading difficulties in Chinese inclusive classrooms. This study examined the effectiveness of ReadVision, an AI-generated visual scaffolding designed to support text processing through semantically aligned images. A quasi-experimental study with repeated measures was conducted with 146 third-grade students with reading difficulties across 21 inclusive classrooms. Results showed that students receiving the intervention demonstrated significantly greater improvements in overall reading comprehension than their peers, with particularly strong gains in inferential comprehension. Although gains attenuated slightly after the intervention, a meaningful advantage was maintained at delayed post-test. These findings provide classroom-based evidence that AI-generated visual scaffolding, when designed as instructional supports, can enhance reading comprehension among students with reading difficulties and underscore the potential of AI-supported visual interventions for inclusive and equitable literacy instruction.