ENHANCING SPEAKING PROFICIENCY OF UZBEK EFL STUDENTS THROUGH GENERATIVE ARTIFICIAL INTELLIGENCE: CHALLENGES, STRATEGIES, AND FUTURE DIRECTIONS

Abstract

Speaking remains one of the most challenging skills for English as a Foreign Language (EFL) learners in Uzbekistan due to limited exposure to authentic communication and large class sizes. This study investigates the effectiveness of generative AI tools (ChatGPT, Grok, ELSA Speak, and Google Gemini) in improving speaking proficiency among Uzbek university students. A quasi-experimental mixed-methods design was employed with 160 second-year students from three major universities in Tashkent. The experimental group (n=80) engaged in AI-supported speaking practice for 16 weeks, while the control group (n=80) followed traditional methods. Results showed a significant improvement in speaking fluency, pronunciation accuracy, and confidence in the experimental group (p < 0.001). Qualitative findings highlighted increased willingness to communicate and reduced speaking anxiety. However, issues such as AI hallucination, lack of cultural nuance, and unequal access to technology were identified. The study proposes a practical framework for integrating generative AI into speaking instruction in the Uzbek context and offers recommendations for educators and policymakers.