International Journal of
Information and Education Technology

Abstract

As Artificial Intelligence (AI) reshapes teaching and learning, the application of Generative AI (GenAI) to the design of literature learning materials offers a novel way to help teachers overcome limitations of visualization and artistic expression. This study proposes VisualLit, a context-aware GenAI framework that generates illustrative images from curriculum-grounded reference data while maintaining the consistency of characters, spatial contexts, and symbolic detail across image sequences, and that screens outputs through an Ethical–Pedagogical Validation Layer (EPVL). The framework was deployed across 15 sixth-grade Literature lessons using two image-generation back-ends: OpenAI’s DALL-E3 and Google’s Nano Banana (Gemini 2.5 Flash Image). In a multi-site evaluation, 17 literature teachers from five secondary schools across two provinces completed a blind preference comparison and a usability survey, and 79 sixth-grade students at a single school took part in a pre-test/post-test quasi-experiment. Teachers preferred the reference-based (curriculum-grounded) outputs across all five evaluation criteria, most strongly for character consistency (88.2%), and rated the system favorably across five usability dimensions (overall mean = 4.5/5.0, Cronbach’s α = 0.84), with aesthetic quality of images (mean = 4.73) and ethical-pedagogical standard (mean = 4.65) rated highest. In the classroom study, students taught with VisualLit materials showed large gains in event-detail recall, symbolic appreciation, and overall achievement (Cohen’s d = 0.91–1.50 for cognitive outcomes; d = 1.35 for affective outcomes; 95% CIs). The study nonetheless affirms the potential of GenAI to innovate literature education, supporting students’ creative capacities, multimodal learning, and humanistic spirit in alignment with the orientation of the 2018 Vietnam General Education Curriculum.