Abstarct: This study examines the impact of AI-driven writing feedback tools, particularly Grammarly and ChatGPT, on developing grammatical accuracy among Iraqi university EFL students. Using a qualitative instrumental case study design, the research followed two undergraduate participants representing different proficiency levels over a four-week intervention in a public university context. Data were collected through screen recordings, think-aloud protocols, stimulated recall interviews, background information, and textual analysis of initial and revised argumentative essays. The findings reveal a noticeable reduction in surface-level grammatical errors, especially in subject–verb agreement, article use, tense consistency, and copula omission. Grammarly proved especially effective in identifying and correcting immediate mechanical and sentence-level errors, whereas ChatGPT encouraged deeper mextalinguistic reflection through explanation, dialogue, and contextual clarification. Across the intervention, the participants showed a gradual shift from blind acceptance of automated suggestions toward more informed correction and critical evaluation, indicating increased awareness, confidence, and engagement. The study concludes that AI feedback tools can serve as valuable pedagogical scaffolds for improving grammatical accuracy when they are integrated into guided instructional practice and complemented by teacher mediation rather than treated as replacements for human feedback.