Abstract

The growing integration of artificial intelligence (AI) in education has expanded opportunities for interactive, technology-mediated language learning. This study investigates the influence of AI-mediated speaking practice using ChatGPT-5 on Modern Standard Arabic (MSA) speaking fluency and willingness to communicate (WTC) among beginner Arabic as a Foreign Language (AFL) learners (A1–A2). It also examines the linguistic and technical challenges learners encounter during AI-supported speaking interaction. Using a mixed-method case study design, the study involves five beginner AFL learners. Data are collected through pre- and post-speaking tests, a modified online WTC questionnaire, semi-structured interviews, and systematic classroom observation. Speaking fluency is analyzed using Skehan’s (2003) utterance fluency framework, focusing on speed fluency, breakdown fluency, and repair fluency. Findings reveal overall positive trends in learners’ speaking performance following AI-mediated practice. The most consistent improvements occur in speed fluency and breakdown fluency, as participants speak faster and produce fewer unfilled pauses. In contrast, repair fluency shows less stable and more individualized development, reflecting the multidimensional and non-linear nature of fluency growth. Results also indicate increased willingness to communicate in MSA, linked to greater confidence, reduced fear of negative evaluation, and repeated experiences of being understood despite linguistic inaccuracies. Qualitative evidence suggests that some learners transfer interactional strategies developed through AI practice—such as clarification requests and meaning negotiation—into authentic communicative contexts. Despite these benefits, challenges remain. Linguistically, ChatGPT-5 sometimes produces vocabulary beyond learners’ proficiency levels and provides limited pronunciation-focused feedback. Technically, participants report issues related to generic or inaccurate feedback, fast speaking pace, interruptions, and inconsistent retention of user instructions, which may disrupt interactional flow and increase cognitive load. The study suggests that AI-mediated speaking practice is most effective when used as a supplementary tool alongside face-to-face instruction, supported by structured tasks and guided feedback. Future research should explore longer interventions, larger samples, and diverse proficiency levels and Arabic varieties.

School

School of Humanities and Social Sciences

Department

Applied Linguistics Department

Degree Name

MA in Teaching Arabic as a Foreign Language

Graduation Date

Spring 6-15-2026

Submission Date

2-10-2026

First Advisor

Dr. Raghda El Essawi

Second Advisor

Dr. Zeinab Taha

Third Advisor

Dr. Shahira Yacout

Committee Member 1

Dr. Raghda El Essawi

Committee Member 2

Dr. Zeinab Taha

Committee Member 3

Dr. Shahira Yacout

Extent

170p.

Document Type

Master's Thesis

Institutional Review Board (IRB) Approval

Approval has been obtained for this item

Disclosure of AI Use

Thesis text drafting; Thesis editing and/or reviewing

Other use of AI

NA

Download

Share

COinS