Abstract
The increasing use of generative AI chatbots in writing education has generated growing research interest; however, limited attention has been given to their application in writing skills development across different languages and learning contexts. To address this gap, this systematic review analyzed empirical studies on the use of AI chatbots to support writing skills. Guided by PRISMA protocols, 48 studies published between 2023 and mid-2025 were systematically analyzed to identify publication characteristics, research designs, and technologies used. The findings revealed a rapid increase in research interest, with ChatGPT emerging as the most frequently examined tool and undergraduate EFL students as the most common participant group. Most studies employed mixed-methods approaches implementing a single chatbot to investigate learner perceptions. Chatbots were primarily utilized as a conversation partner and feedback provider, while their potential as an evaluator or resource provider was underexplored. In addition, insufficient reporting of research contexts and participant characteristics limited the interpretability of several studies. This review identifies methodological and design limitations, highlights gaps in the existing literature, and provides empirical insights to inform future research on AI-assisted writing education.