<![CDATA[This study compared three widely used educational image generators, Gemini, ChatGPT, and Canva AI, to examine their suitability for teaching Arabic to non-native speakers, in light of the limited criteria that regulate cultural appropriateness and cognitive effectiveness of AI-generated images in a language characterized by morphological and visual density like Arabic. The study relied on an evaluation scale guided by the cognitive theory of multimedia learning, and also relied on the judgments of eight experts in Arabic linguistics, educational technology, and instructional media design. The images generated from a single uniform prompt were evaluated according to three dimensions: linguistic and cultural accuracy, visual clarity, and pedagogical appropriateness for oral and written expression tasks. The descriptive and thematic analysis showed a clear progression in multimodal scaffolds; Gemini provided the most consistent images and those closest to cultural plausibility, outperforming ChatGPT and Canva AI in reducing external visual complexity, providing indicators that supported idea generation, activating vocabulary, and organizing discourse. The study also presented a practical framework that can benefit teachers and researchers in evaluating AI-generated educational images, highlighting linguistic and cultural fidelity as a central criterion in AI-supported multimodal design.]]>
Copyrights © 2026
