Adaptive versus Fixed Metacognitive Prompting in GenAI-Assisted EFL Writing: Effects on Self-Regulated Learning and Writing Performance

This study tested adaptive metacognitive prompting (AMP) against fixed metacognitive prompting in a generative artificial intelligence (GenAI)-assisted English as a Foreign Language (EFL) writing environment. Seventy-seven Chinese EFL undergraduates were randomly assigned to an adaptive prompting group ( n = 39) or a fixed-prompt control group ( n = 38) in a pretest-posttest-delayed posttest design. Both groups used the same GenAI writing interface. The adaptive group received need-triggered prompts based on behavioral and text signals. Mixed ANOVAs, simple effects tests, revision analysis, and lag sequential analysis were applied to questionnaire data, writing scores, and interaction logs. Adaptive prompting produced a short-term posttest advantage in total self-regulated learning (SRL), higher overall writing scores, and more monitoring and revision-oriented behavior during supported writing. However, this advantage was not maintained at the delayed posttest, and the writing-score interaction was not significant. Sequential analysis showed a prompt-supported monitoring-to-reflection transition under adaptive prompting. The findings suggest that contingent metacognitive prompts strengthen regulation during GenAI-assisted writing, although the evidence supports scaffolded short-term performance rather than durable skill acquisition.