The integration of Artificial Intelligence (AI) into language education has opened new avenues for enhancing academic writing instruction, particularly when combined with human-mediated pedagogical frameworks. This study investigates the impact of AI-assisted feedback via ChatGPT paired with teacher-mediated Dynamic Assessment (TMDA), referred to as AI-DA synergy—on the academic writing development of Iranian nursing students who enrolled in a university-level English as Foreign Language (EFL) course. Grounded in Vygotsky’s Sociocultural Theory (SCT), the study aims to evaluate both the effectiveness of AI-DA synergy and learners’ perceptions of its implementation. Employing a mixed-methods design, the research involved 60 B1-level nursing students randomly assigned to three groups: AI-DA, DA-only, and a control group, each comprising 20 learners. Participants completed two CEFR B1-aligned timed descriptive writing tasks in pretest and posttest phases, yielding quantitative measures of their writing proficiency. Qualitative data were obtained via semi-structured interviews to explore students’ writing development and their critical reflections on AI-DA feedback. ANOVA analyses revealed that the AI-DA group significantly outperformed both comparison groups in writing proficiency. Thematic analysis of student perceptions highlighted five key themes: the necessity of teacher mediation, emotional engagement, skill development, limitations of AI assistance, and trust negotiation. Students expressed appreciation for the dual support system, noting that teacher feedback provided contextual clarity and emotional reassurance, while AI offered immediate, personalized suggestions. Findings underscore the transformative potential of AI-DA synergy in enhancing learners’ academic writing through personalized, scaffolded feedback, urging educators to use AI with human mediation to foster deeper engagement, contextual clarity, and emotionally responsive instruction.