Abstract
<jats:p>Rapid emergence of generative artificial intelligence (AI), is transforming the landscape of writing instruction. While current applications often emphasize surface-level uses like grammar correction or paraphrasing, this study proposes a deeper pedagogical integration: conceptualizing AI as metacognitive and dialogic scaffold capable of supporting higher-order thinking. Drawing on Cognitive Load Theory, Sociocultural Theory, Frameworks of Dialogic Learning, and Critical Pedagogy—the research situates AI as a non-human dialogic agent that can interactively engage students within their Zone of Proximal Development. Using a Design-Based Research methodology, the study implemented an AI-integrated instructional model in the course of English as a Second Language (ESL). The intervention combined Socratic instruction in strategic prompting with guided use of AI across writing stages. Collected data and findings indicate that AI, when framed as Socratic dialogue partner, can reduce extraneous cognitive load and increase engagement with germane cognitive processes, particularly in ideation and revision. Students reported greater clarity in argument structure and enhanced rhetorical awareness. Importantly, personalized AI feedback was perceived as a motivating guide contingent on students’ ability to co-direct the interaction through intentional prompting strategies. The study contributes a theoretical and pedagogical framework for AI-responsive writing instruction and raises critical questions about ethical literacy and the evolving role of educators in AI-augmented learning environments.</jats:p>