Using Gemini for Formative Assessment in English Academic Writing - Critical Insights into The AI Tool’s Efficacy

The emergence of Artificial Intelligence (AI) has triggered revolutionary transformations in language teaching and learning. When it comes to academic writing, current educational practitioners must more than once wonder which AI-powered tools, among the overwhelming number mushrooming recently, can assist their learners’ self-study by providing reliable and relevant feedback. This paper explores the effectiveness of Gemini, a large language model (LLM) developed by Google AI, in providing rubric-aligned commentary on student essays. The article employed a mixed-methods approach in which quantitative data are collected from academic writing samples while qualitative data are coded from Gemini-assisted feedback. Through the critical analysis of the comments provided by Gemini on twenty students’ essays, against the IELTS Writing Task 2 band descriptors, Gemini’s feedback tends to be more consistent when it comes to task achievement and coherence and cohesion, with rubric or band descriptors included in the prompt. Within each criterion in the rubric, the initial indicators tend to be more adequately examined. Also, paragraphing, spelling, and punctuation are the indicators that are neither consistently nor sufficiently commented on. These findings lay a foundation for language educators to evaluate the efficacy of LLM-assisted learning tools in academic writing education, paving the way for their proper application in classroom instruction.