Comparing Listening Gains from AI Text-to-Speech vs. Real Human Voice Recordings in Uzbek EFL

This study investigates the comparative effectiveness of AI-based text-to-speech (TTS) technology and traditional human voice recordings in fostering listening comprehension among Uzbek EFL learners. With the growing implementation of AI in education, TTS tools offer a promising alternative to human recordings, particularly in resource-limited contexts. Seventy-two intermediate-level secondary school students in Tashkent participated in a five-week intervention, where one group listened to AI-generated audio while the other engaged with native-speaker voice recordings based on identical scripts. Both groups completed comprehension quizzes and summary tasks to measure improvement. Findings indicate that while both approaches led to significant listening gains, students exposed to human voices performed slightly better in interpreting emotional tone, stress, and implicit meaning. Meanwhile, the AI TTS group demonstrated more consistent progress in recognizing vocabulary and understanding explicit content. Learners appreciated the clarity, predictability, and slower pacing of the TTS voices, which lowered anxiety and improved focus. Despite some limitations in expressiveness and naturalness, AI TTS tools proved to be an effective supplemental resource, especially in classrooms where access to native-speaker recordings is limited. The study concludes that a hybrid approach combining TTS and human voice recordings could offer the most balanced and accessible strategy for EFL listening instruction in Uzbekistan.