Enhancing ESL Students’ Speaking Fluency and Accuracy with AI-Integrated Learning: An Experimental Study

Abstract

This study explores how integrating AI into learning can enhance English language skills among ESL learners, particularly in speaking. It examines the efficacy of various AI- powered tools that improve the speaking proficiency of ESL students at COMSATS University Lahore Campus. Based on the TPACK framework and Krashen’s Input and Affective Filter hypotheses, this study employs a single-group, mixed-methods embedded experimental design. An eight-week use of integrated AI-based technologies like ChatGPT, TalkPal AI, Speechace, and Smalltalk2Me was carried out at COMSATS. A total of seventeen undergraduate and postgraduate students took a speaking test based on IELTS speaking rubrics before and after the intervention and participated in structured interviews to gauge perceptions of the effectiveness of various AI-powered language learning tools in providing practice in English speaking. Results obviously indicate overall statistical gains in speaking. In particular, improvements are remarkable in areas of fluency, pronunciation, and grammatical accuracy. Qualitative semi-structured interviews revealed that motivation and confidence in practicing English were greater with immediate feedback, provided individually through AI-enabled tools in a low-stress practice environment. These are followed by concrete suggestions as to how language teachers, curriculum designers, and policymakers can apply intelligent technologies as part of planning higher education courses in languages, noticing an improvement in vocabulary learning and in learners' confidence, with a view toward greater access to language-practice opportunities in the general higher-education sphere.