Analyzing Composition Complexity in Revised Versions through Artificial Intelligence-Based Writing Assistants

Artificial intelligence-based correction systems have brought convenience to English learners’ writing practice. Grammarly is an online AI-powered essay revision software. This paper explores the changes in the writing complexity of essays after being corrected by this software compared to their original versions. Lexical complexity and syntactic complexity are important aspects of second language writing. Therefore, this study selects three lexical complexity indicators and ten syntactic complexity indicators to measure the changes in essays before and after being corrected by Grammarly. The research findings reveal that university-level English essays, after being modified by the AI-powered correction system, do not exhibit significant increase in lexical aspects, specifically in lexical density and lexical sophistication, while the indicator of lexical variation shows a noticeable decrease. On the other hand, in terms of syntactic complexity, four indicators—MLC(mean length of clause), CPC(coordinate phrase per clause), CP/T(coordinate phrases per T-unit), and CN/C(complex nominals per clause)—demonstrate a significant increase. This suggests that students can benefit from Grammarly’s assistance in essay correction training in terms of syntactic complexity, while the improvements in vocabulary complexity are not substantial.