From Synthetic Fluency to Authentic Learning: Redesigning ESL Teaching and Assessment Methodologies in the Age of AI

Abstract

This paper emerges from a practical concern observed in ESL classrooms: students are producing increasingly polished written work with AI support, yet struggling to demonstrate the same level of competence under exam conditions. This gap raises important questions about how teaching, learning, and assessment currently align. Recent studies indicate significant discrepancies between AI-supported coursework and performance in proctored examinations, raising concerns about the validity of traditional assessment practices. This paper conceptualises this phenomenon as synthetic fluency, where students produce linguistically accurate outputs without corresponding internalised competence.

Focusing on ESL provision within General Foundation Programme (GFP) in Oman, the study argues that the core issue is not academic misconduct, but a systemic failure of pedagogical alignment. Drawing on recent international research and practice-based insights and based on the analysis of GFP results for summative assessments in different settings, the paper analyses how AI has disrupted the relationship between teaching methodologies, learning processes, and assessment systems.

In response, the study proposes an integrated framework for redesigning ESL pedagogy in AI-augmented environments. We can reconceptualise teaching as a process of cognitive apprenticeship supported by guided AI use, while assessment will shift from product-based evaluation to process-based validation through interconnected tasks, AI transparency, and oral defence mechanisms. The paper introduces a three-layer model consisting of AI-declared work, cognitive trace, and authenticity validation.

This model can be realistically implemented in GFP contexts, with direct implications for how we design assessment, support learning, and define academic integrity.