Leveraging GenAI for Automated Feedback on Source‐Based Writing: A Three‐Stage Model

Abstract Writing from multiple sources is a critical skill that requires both reading and writing abilities. It involves comprehending and evaluating source texts while synthesizing information across them. Due to these demands, feedback on such writing typically relies on teachers, who must consider the source texts during evaluation. Although technological tools have been introduced to support teacher feedback, existing automated systems often inadequately assess source‐based writing, as they focus solely on students' final texts and neglect the source texts, leading to invalid assessments. The advent of generative artificial intelligence (GenAI), capable of understanding and analyzing large volumes of text simultaneously, offers a promising solution. In this article, we propose a three‐stage GenAI‐enabled feedback model for source‐based writing. This model aligns with the key stages of the source‐based reading‐to‐writing process: evaluating individual sources, mapping multiple sources, and synthesizing information across sources. We explore how GenAI can be incorporated into each stage to provide personalized, targeted feedback, helping students navigate the complexities of source‐based writing. To illustrate the application of this model, we focus on literature review writing, a specific type of source‐based writing commonly found in English for Academic Purposes classrooms. By offering this model, the article provides educators with a practical framework for leveraging GenAI to improve students' ability to write from multiple sources.