Sequencing feedback at scale: A cluster-randomized trial comparing AI-powered AWE and teacher feedback in Indonesian EFL academic writing

Large Indonesian EFL writing classes often struggle to provide timely, detailed teacher feedback. Automated Writing Evaluation (AWE) offers rapid responses, but its role in academic writing remains unclear. This mixed-method study compared AWE-led feedback with teacher feedback in a 10-week Indonesian undergraduate writing course and proposed a Hybrid AWE–Teacher Feedback Model. A PRISMA-guided review informed feedback design. The empirical phase used a cluster-randomized trial (N = 88; four classes): two received AWE feedback (Grammarly), and two received teacher feedback. Writing quality was measured by pre/post essays, and engagement was measured by revision logs, questionnaires, and interviews. Both groups improved. AWE produced greater gains in language use, vocabulary, and revision frequency, and higher post-test scores. Teacher feedback contributed more to idea development and rhetorical clarity. The results point to a staged strategy: use AWE for early, frequent revision, and reserve teacher feedback for higher-order concerns in writing.