Authentic Tourism Tasks + AI Speaking Studio: A Scalable Model for Fluency, Intelligibility, and Transfer

Abstract

This article reported the design and early implementation of a six-week Task-Based Language Teaching cycle integrated with an AI-assisted “Speaking Studio” for third-semester Sustainable Tourism students (n=20; 10 hours/week) in Palenque, Mexico, conducted in partnership with Hotel Misión Palenque. The aim was to improve fluency, intelligibility, and the transfer of classroom English to authentic service encounters. The intervention combined two authentic deliverables—a 90-second concierge pitch and a bilingual micro-artifact—with a weekly routine of deliberate practice (shadow–record–reflect) supported by ChatGPT to generate genre models, phrase banks, clarity checks, pronunciation targets, and retrieval micro-quizzes. Continuous formative assessment (exit tickets, error logs) and a summative four-band rubric (intelligibility, fluency, appropriacy, interaction) were planned, complemented by transcript-derived analytics (words per minute, mean utterance length, lexical variety), partner evaluations, and a short speaking-anxiety scale. Expected outcomes included faster speech with stable accuracy, improved stress and service-move control, higher confidence when engaging visitors, and the adoption of at least one student-produced artifact by the hotel. The model proved scalable, low-cost, and replicable, easing the feedback bottleneck without replacing teacher judgment.