GenAI-supported adaptation of children’s storybooks into vocabulary materials in Indonesian primary education

Abstract

Generative Artificial Intelligence (GenAI) is increasingly used to support the adaptation of children's storybooks into English as a foreign language vocabulary material in Indonesian primary education, yet limited evidence explains how such use shapes vocabulary, cognitive, multimodal, and cultural aspects of material quality. This qualitative multiple-case study examined three areas: (1) vocabulary quality, cognitive demand, multimodal coherence, and cultural relevance of children’s storybooks and teachers' existing GenAI-generated vocabulary materials; (2) the prompt components, design problems, and teacher revision patterns emerging from comparisons of storybooks, standard GenAI outputs, curated GenAI outputs, and teacher revisions; and (3) teacher curation and parent mediation in GenAI-supported storybook-based vocabulary practices. Participants included primary teachers, parents, and supporting school informants, while the corpus comprised 32 children’s storybooks and related GenAI artifacts. A comparative analysis matrix was used to examine vocabulary quality, cognitive demand, multimodal coherence, and cultural relevance. Artifact documentation and walkthrough protocols traced prompt-output-revision patterns, while interviews and focus group discussions examined teacher curation, parent mediation, and classroom/home enactment. Data were analyzed using directed qualitative content analysis and reflexive thematic analysis. Within this bounded qualitative dataset, standard GenAI outputs tended to reduce vocabulary specificity, shift tasks toward factual recall, weaken multimodal coherence, and introduce culturally generic references. Curated GenAI outputs showed stronger alignment with target vocabulary, inferencing demands, local contexts, and classroom usability. The findings suggest that GenAI is most useful as a drafting tool within teacher- led curation, culturally responsive prompting, human-in-the-loop review, and AI literacy support.