This study examines the impact of AI-assisted language learning tools on EFL learners' lexical choices, focusing on lexical diversity, lexical sophistication, and contextual appropriateness. Despite the growing integration of AI in language education, little research has examined how AI shapes learners' actual lexical selection during production. To address this gap, the study adopts a longitudinal, quasi-experimental mixed-methods design, combining a quantitative pretest-posttest, control-group framework with qualitative thematic analysis to capture both measurable linguistic shifts and learner perceptions. Between 60 and 80 intermediate-to-upper-intermediate EFL learners will be randomly assigned to experimental and control groups. Data are collected through writing tasks, discourse completion tasks, questionnaires, and semi-structured interviews. Learners exposed to AI are expected to show greater lexical diversity, richer vocabulary, and more contextually appropriate choices than those receiving conventional instruction, with implications for EFL teaching, curriculum design, and educational policy.
