Navigating Between the Scylla and Charybdis of Hype and Fear: Decolonizing ELT Materials Through Generative Artificial Intelligence

Abstract Despite calls from critical applied linguists to decolonize ELT materials that reproduce colonial ideologies, teachers working under neoliberal constraints and attendant time pressures often have to deliver textbook content as is rather than revising it. This study explores to what extent a state‐of‐the‐art large language model (LLM) can work as a practical tool for decolonizing materials. We focus on a textbook unit widely used in Hong Kong that reproduces a white savior narrative and stereotypical representation of Africa and Africans as a telling case of coloniality. The unit was processed through an LLM with a decolonizing prompt, and both the original and AI‐revised versions were subjected to critical discourse analysis to understand how social actors and actions are represented and constructed at the macro and micro levels. Findings suggest that while the revision comes with significant limitations, the LLM decolonized the materials to an extent by reframing representations, redistributing agency, and refining word choice. While contending that AI has the potential to inject much‐needed agency back into the language teaching profession, we also argue that using AI has significant implications for teacher education, not only in terms of developing critical consciousness and critical AI literacy but also of teachers' knowledge base.