Generative AI-enabled writing feedback in primary school English instruction: application landscape, practical dilemmas, and a teacher-led collaborative framework

Writing feedback is a pivotal yet chronically constrained component of primary school English writing instruction, squeezed by large class sizes, low marking frequency, and homogeneous comments. Generative Artificial Intelligence (GenAI), with its immediacy, personalization, and dialogic interaction, opens new possibilities for reconstructing writing feedback. Drawing on a narrative synthesis of 52 Chinese and international sources, and addressing a literature in which research on AI writing feedback concentrates on higher education and on the learner side, this paper maps the application landscape of GenAI-enabled writing feedback in primary school English, consolidates and examines five practical dilemmas concerning accuracy, appropriateness, teacher literacy, ethics, and evaluation orientation, and proposes a Teacher-led, AI-assisted, Student-centered Collaborative Feedback Framework (TAC-F). The framework assigns teachers responsibility for goal setting, affective support, and feedback verification, positions GenAI as an initial assessor and material provider, and treats students as reflective revisers within a pre-writing, while-writing, and post-writing closed loop, supported by prompt-template libraries, feedback-verification routines, school-based teacher AI literacy development, and ethical data norms. Preliminary multi-tool observations, together with a six-dimension observation framework and a same-prompt pilot protocol, are developed to help teachers screen AI-generated feedback; the observations suggest that current tools are reliable on explicit errors but unstable at the discourse level. Six implementation suggestions are advanced to inform the digital transformation of primary school English writing instruction.