Pedagogical viability of generative AI voice in English language teaching

Abstract

The integration of generative artificial intelligence in the production of educational materials raises questions about its pedagogical viability in contexts with institutional constraints. This article documents the experience of a public upper secondary education institution that, faced with the unexpected absence of an audio component in an English language textbook series, used generative AI voice tools to produce the required materials before the start of the school year. The study adopted a mixed-methods approach with quantitative predominance, combining a retrospective reconstruction of the production process with the administration of a 15-item Likert questionnaire to 47 teachers with prior experience using traditionally recorded audio materials. Results indicate a favorable and consistent perception across the four blocks of the instrument (M = 3.98, α = 0.972), with pedagogical value as the highest-rated dimension and curricular coherence as the area of greatest consensus. Qualitative analysis of an open-ended question confirmed the resource's usefulness for teaching practice, while the limitations identified by teachers pointed to infrastructure conditions rather than audio quality. Findings suggest that generative AI voice technology represents a viable alternative for the institutional production of audio materials when implemented within a defined pedagogical framework and teacher evaluation process, with relevant implications for institutions facing similar constraints across Latin America.