AI-Based Writing Assessment in Second Language Context: A Bibliometric Analysis of Trends in Coherence and Cohesion

The use of AI-based writing assessment in language education is growing rapidly, offering both opportunities and challenges in evaluating second language writing (L2). While these tools are increasingly common, most still focus on surface-level accuracy, such as grammar and spelling, and often overlook discourse-level elements like coherence and cohesion. These features are essential for L2 learners to produce texts with logical flow and organization. This study investigates research trends in AI-driven assessment of L2 writing from 2020 to 2025 through a bibliometric analysis of 332 articles retrieved from databases such as Scopus, JSTOR, ScienceDirect, and Emerald Insight. The analysis followed five key steps: setting objectives, selecting and cleaning data, performance analysis, keyword co-occurrence mapping, and interpretation. VOSviewer software was used to visualize thematic patterns and identify research gaps. Findings show a significant increase in AI-related research within language learning, yet coherence and cohesion remain underrepresented. Most studies emphasize technological aspects, such as system efficiency, tool development (e.g., ChatGPT), and AI innovation, rather than deeper writing features. The keyword analysis confirms this imbalance, highlighting a disconnect between research focus and educational priorities. This study calls for closer collaboration among educators, linguists, and AI developers to ensure assessment tools align with writing pedagogy. By mapping existing research trends and revealing neglected areas, this study contributes to improving AI-based writing assessment tools to better support meaning-making, structure, and clarity in second language writing.