BEYOND THE PROMPT:

Abstract

The post-2022 paradigm shift in educational technology has transitioned Artificial Intelligence from a basic grammar-checking tool into an active partner in "generative authorship." This evolution challenges traditional notions of student agency, assessment integrity, and text ownership in English as a Foreign Language education. This paper presents an evidence-based pedagogical framework that merges deep cognitive learning theories with empirical qualitative data to address this disruption. Theoretically, the study adapts Vygotsky's Zone of Proximal Development, Cognitive Load Theory, and Self-Determination Theory to re-evaluate the human-AI collaborative writing loop. It frames the central tension as a choice between AI acting as a cognitive scaffold that aids developmental growth or a cognitive bypass that masks underlying learning gaps with sterile, algorithmic text perfection. Empirically, the paper draws on a Reflexive Thematic Analysis exploring university teachers' confidence, anxieties, and trust regarding the assessment of AI-mediated writing. The qualitative findings reveal significant fractures in traditional grading criteria, driven by an inability to distinguish developmental, "approximate language" from automated generation, which severely undermines teacher assessment confidence. To bridge this divide, this study proposes an Adaptive Writing Ecosystem that moves beyond screen-level prompt engineering to prioritize the internal cognitive and affective states of the writer. The framework advocates for a structural shift from product- to process-based rubrics, institutional guidelines separating brainstorming from drafting phases, and the integration of oral verification protocols. Ultimately, the study concludes that while technology personalizes the text, only humanized pedagogy can personalize the meaning-making process.