<![CDATA[This PRISMA 2020 systematic review maps AI technologies, populations, and speaking-related outcomes, evaluates comparative pedagogical evidence, and identifies gaps for a future research agenda. Framed by PICo with SPIDER elements, peer-reviewed English-language articles 2020–2025 were retrieved from Scopus and ERIC, with hand-searching in IEEE Xplore, ACM Digital Library, and Web of Science. Two reviewers screened records independently, disagreements were resolved by a third. Quality was appraised using the JBI Critical Appraisal Tools and the Mixed Methods Appraisal Tool (MMAT), synthesis followed a narrative approach. Of 1,485 records, 19 studies met inclusion criteria. Reclassified pedagogically, 14 of 19 were technical or model-development studies, 2 of 19 qualitative case studies, and 3 of 19 diagnostic, descriptive, or adoption studies. Speech and audiovisual systems dominated the evidence base 11 of 19, followed by NLP text-only applications 4 of 19 and a heterogeneous cluster 4 of 19. Only 4 of 19 studies involved non-native learners, none operationalised fluency, prosody, comprehensibility, or communicative competence. No study directly compared AI-supported and non-AI Arabic instruction. The Arabic AI literature demonstrates technological feasibility but not pedagogical superiority. Second Language Acquisition grounding remains limited, implicit, or insufficiently operationalised.]]>
Copyrights © 2026
