Generative AI and media accessibility for disabled audiences in Ibero-America

Abstract

Generative artificial intelligence (GenAI) can now produce, at marginal cost, the services that make media accessible to audiences with disabilities: automatic captioning and subtitling, audio description, sign-language rendering, text simplification, and image description. In the underfunded, platform-dependent, and linguistically plural media ecology of Ibero-America, this capacity is widely promoted as the means to close a persistent accessibility gap. Yet the same systems reproduce ableist assumptions and fail silently, raising the prospect that automation will scale a counterfeit of access rather than access itself. This conceptual article asks under what conditions GenAI expands, rather than simulates, communicational accessibility for persons with disabilities. Through a critical synthesis of media accessibility studies, the disability-and-AI literature, and regional communication research, with every source verified at the level of its digital object identifier, it constructs a four-dimensional framework (provision, fidelity, representation, and governance) positioned along two axes that distinguish scale from situatedness and substitution from augmentation. Applied to the principal access services, the framework shows that no service is uniformly safe and that governance is the consistent variable separating genuine access from accessibility theatre. The article concludes that the decisive factor is not the capability of the model but the governance of its deployment, and it proposes a participatory, empirical, and policy agenda in which the region's public-interest media could lead.