Abstract The application of generative artificial intelligence(AI)in academic writing has been driving a transformation in traditional academic writing instruction.However,little research has been done on the ways in which AI assists in academic learning.Lexical bundles,as key components of academic writing,are defined as recurrent multi-word sequences and serve as distinguishing feature of discourse across genres,and disciplines.Adopting a corpus-based approach,this study examines the convergent and divergent usage of 4-word lexical bundles between AI-generated and scholar-written English abstracts in international chemistry journal articles.The results show that AI relies more heavily on lexical bundles but exhibits limited bundle diversity in constructing chemistry abstracts,while does not conform to the English academic writing convention that dominated by of-phrases.Structurally,AI prefers clausal bundles,while scholars favor phrase-based bundles.Functionally,both AI and scholars prioritize research-oriented bundles.Nevertheless,AI overuses text-structuring bundles while under-utilizing framing and stance bundles.These limitations reflect the inherent constraints of generative AI pretraining models,which lack personalization and the rhetorical sophistication required for effective academic persuasion.This study underscores the importance of raising chemistry scholars' awareness of lexical bundles and advancing AI-human collaboration of academic English writing instruction in the field of chemistry.
YANG Chuan-Ming, GUO Yun-Jie, LENG Li-Tian. Comparative Study on Lexical Bundles in AI-Generated and Scholar-Written Abstracts of Chemistry Journal Articles[J]. Chinese Journal of Chemical Education, 2025, 46(8): 93-102.
