Generative AI as a Scaffold in EFL Writing: A Correlational Study of Writing-Process Mediation among Algerian Master's Students

Abstract

The rapid integration of generative artificial intelligence (GenAI) into higher education justifies this study, as it increasingly shapes EFL writing, learner autonomy, cognitive engagement, and authorship. Grounded in sociocultural theory, writing process theory, cognitive load theory, and AI literacy, the study builds on previous CALL and AI-assisted writing research while addressing limited evidence from Algerian higher education. It aims to examine how Algerian Master’s students use GenAI across EFL writing stages and whether this use reflects scaffolded or substitutional mediation. Its novelty lies in its stage-based correlational approach and the proposed scaffold–substitute continuum. A cross-sectional descriptive-correlational design was adopted, using a questionnaire completed by 549 students and 312 usable AI-use logs. Quantitative data were analyzed through descriptive statistics and Spearman’s correlation, while qualitative logs were examined thematically. Findings show that GenAI use was highest during revising and editing, mainly for lexical refinement, grammatical correction, organization, and revision support. Three profiles emerged: scaffolded, hybrid, and substitutional use. The study concludes that GenAI can support EFL writing when used critically, but may reduce cognitive engagement and authorship awareness when used passively. Theoretically, it contributes a scaffold–substitute continuum; practically, it highlights the need for AI literacy, process-oriented assessment, and ethical authorship practices.