USING ARTIFICIAL INTELLIGENCE TOOLS FOR AUTOMATED TEXT DIFFERENTIATION IN ENGLISH LESSONS

Öz

This article aims to develop a theoretically grounded, criteria-based framework for evaluating the quality of AI-generated text adaptations in EFL middle school settings. The study is theoretical in nature and employs a systematic critical literature review of 20 peer-reviewed sources (2023-2026), supplemented by comparative analysis and theoretical synthesis. Three analytical procedures were applied: comparative reading of selected sources, classification of research gaps, and deductive theoretical integration. The central finding is a unified three-criterion evaluation framework, comprising difficulty alignment, readability alignment, and differentiation completeness, which is derived from the integration of the Zone of Proximal Development by Vygotsky L. S., the Input Hypothesis by Krashen S. D., and the Differentiated Instruction model by Tomlinson C. A., and operationalized through CEFR level descriptors and Flesch-Kincaid readability scores. Comparative analysis of four key empirical studies confirmed that no prior work has united all three theoretical frameworks, addressed secondary school learners, or proposed a replicable evaluation methodology in this domain. The proposed framework offers EFL teachers in multilevel classrooms a principled quality assurance procedure for assessing AI-generated materials prior to classroom use, while providing a structured foundation for the empirical phase of research.