English Language Learning with AI: Proficiency Gains and Learner Experience

Abstract

This study investigates the effectiveness of an AI-assisted language learning platform compared to traditional instruction among first-year Saudi university students. Using a quasi-experimental mixed-methods design, the study assigned 147 students to either an AI-assisted group or a traditional classroom group over a six-week period. Quantitative data from pre- and post-tests (TOEFL ITP) revealed significantly greater gains in the AI group (d = 0.85), even after controlling for baseline proficiency. Regression analysis showed that time spent on the AI platform was a strong predictor of learning gains, with each additional 30 minutes of usage correlating with a 1.8-point improvement. Qualitative data from 30 post-intervention interviews highlighted the perceived benefits of immediate feedback, gamified motivation, and self-paced learning, alongside reported challenges such as streak-related anxiety, accent misclassification, and technical issues. Findings support a blended learning model that uses AI to give students consistent feedback, while keeping teachers focused on guiding learning and supporting students emotionally and socially. The study contributes practical and theoretical insights relevant to policymakers, curriculum designers, and educators seeking to implement AI in EFL contexts.