Equipping L2 Teachers for Linguistic Justice in the Age of Generative AI: A Generative‐Curator Heuristic

ABSTRACT Although generative artificial intelligence (GenAI) can support second‐language (L2) teaching and learning through materials development, feedback, revisions, model texts and translation, generated language may also reproduce standard‐language, raciolinguistic and ableist norms. This conceptual article argues that existing AI literacy and teacher‐education frameworks do not yet fully address this language‐specific problem and offers a pedagogical heuristic to fill that gap. Drawing on AI literacy, critical digital literacy, language ideology, raciolinguistic and accessibility scholarship, we argue that L2 teachers should be prepared to be generative curators who critically select, adapt, contest, reject and redesign generated language while preserving learners’ agency, communicative purposes and linguistic repertoires. This heuristic is designed for pre‐service and in‐service L2 teacher education and focuses on the situated analysis of generated‐language artefacts, including feedback, revisions, model texts and translations. Existing frameworks address evaluation, ethics, inclusion and human agency, but do not consistently translate these commitments into procedures for examining generated feedback, revisions, model texts and translations. In response, the article shifts the unit of analysis from broad competency domains to classroom‐facing language that GenAI produces and that teachers are asked to evaluate, adapt or reject. The heuristic pedagogically extends, rather than replaces, these frameworks through three interconnected capacities: understanding GenAI as a sociotechnical language technology, curating generated texts for linguistic and accessibility justice, and exercising reflective professional judgement. A comparative framework analysis, output‐audit questions, a worked vignette, and teacher‐education activities illustrate these capacities. The article contributes a practicable language‐specific heuristic for identifying when GenAI expands communicative possibilities versus standardising language, restricting participation or displacing learner authority.