Answers are grounded in this paper and cite their sources.
Academic writing is a core competence for learners of English as a foreign language (EFL), yet it remains one of the most demanding skills because it requires the simultaneous integration of content development, idea organization, and linguistic accuracy. Many EFL students consequently devote disproportionate cognitive resources to lower-level language concerns, which may hinder higher-order processes such as coherence building and argument development. Drawing on sociocultural perspectives on cognitive scaffolding and the growing capacity of artificial intelligence (AI) tools to deliver immediate, individualized support, this study investigated whether AI-based cognitive scaffolding could enhance EFL students’ academic writing outcomes and whether such effects generalized across different dimensions of writing performance. A quasi-experimental pretest–posttest design was employed over 14 weeks with 53 first-year non-language-major students at a public university in Vietnam. Students’ writing performance was compared between a group that received AI-based scaffolding and a group that received conventional instruction. The findings showed that AI-mediated scaffolding produced a significant improvement in the experimental group’s overall academic writing performance (gain Δ = 0.81) compared with the control group (gain Δ = 0.55). However, the benefits were selective and domain-specific rather than uniform: the intervention markedly enhanced idea generation and language-related performance but had limited impact on organizational structure and content elaboration. These findings highlight both the strengths and the limitations of AI tools in EFL writing instruction and suggest that AI is most effective when embedded in an integrated pedagogy that combines technological scaffolds with explicit teaching and guided support to address higher-order writing demands.
No takes yet. Share an insight, caveat, or question.
Huy et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: