Constructing CEFR-Aligned English Reading Assessments with AI: A Practical Guide for Thai English Teachers in Higher Education

Abstract

This paper provides practical guidance for developing English reading assessments aligned with the Common European Framework of Reference for Languages (CEFR) with the support of generative AI. Although the CEFR is widely used in curriculum design, instruction, and assessment, many instructors face challenges when translating reading descriptors into assessment tasks that appropriately reflect specific proficiency levels, particularly transitional levels such as B1+ and B2+. AI tools offer new opportunities for generating reading materials and assessment tasks, but recent studies show that CEFR level labels alone do not ensure appropriate alignment. Drawing on language assessment theory and recent research on AI-generated CEFR-aligned texts, this paper proposes a specification-driven approach in which reading test specifications are derived from CEFR-aligned instructional materials and target CEFR descriptors to guide AI-assisted assessment development through a three-stage prompting process consisting of reading test specification development, reading assessment generation, and optional review and revision. For illustrative purposes, a baseline prompting approach based on CEFR-aligned textbook tasks and target CEFR descriptors is also presented to represent a realistic classroom assessment development process. Illustrative applications at the B1, B1+, and B2 levels show how reading test specifications can be incorporated into AI prompts to guide the development of CEFR-aligned reading assessments. The paper concludes with practical guidance for adapting AI-assisted assessment development to classroom contexts while maintaining instructor judgment and alignment with the intended reading construct.