Prompting for justice: generative AI, translanguaging, and the equity imperative in bilingual education

Generative AI (GenAI) is entering bilingual and dual language classrooms faster than theoretical and pedagogical frameworks can respond. Trained predominately on English-dominant data, large language models encode linguistic hierarchies that contradict equity-oriented bilingual education stands for. Drawing on translanguaging theory, critical AI literacy research, and empirical studies of language bias, this paper proposes Equity-Centered Prompting Framework (ECPF). ECPF comprises five principles spanning prompt design, output evaluation, student participation, instructional function, and epistemic authority. Unlike existing critical AI literacy frameworks that address technology broadly, or culturally sustaining pedagogies that do not engage AI specifically, ECPF targets the linguistic structure of GenAI and its consequences for bilingual learners, positioning AI interaction itself as an act of equity-oriented pedagogy. The framework was developed through a four-stage systematic theoretical synthesis: construct identification, literature search and selection, thematic synthesis, and principle formulation and validation based on 20 core sources. The contribution is threefold: a translanguaging-based critique of how GenAI architectures. five actionable principles for equity-centered prompting in actual bilingual classrooms, and a research and policy agenda for linguistically just AI integration.