Lexical Richness and Syntactic Complexity in AI-Generated Texts: A Comparison of ChatGPT and Native English University Students’ Writing

With the rapid development of generative artificial intelligence (AI), increasing scholarly attention has been directed to the opportunities and challenges posed by large language models such as ChatGPT for foreign language education. One major concern is that students may become overly reliant on AI, leading to practices such as ghostwriting and plagiarism. It is therefore essential to identify linguistic evidence that can reliably distinguish AI-generated texts from human writing. This study compares texts generated by ChatGPT with same-topic, equal-length essays written by high-proficiency native English-speaking students, with a particular focus on differences in lexical richness and syntactic complexity. The results show that ChatGPT significantly outperforms human writers in lexical sophistication, lexical diversity, and lexical density, especially in its use of low-frequency vocabulary and complex noun phrases. In addition, ChatGPT demonstrates advantages across multiple dimensions of syntactic complexity. However, it is less effective than native students in handling subordinate structures and in expressing complex logical relationships. The study further discusses how ChatGPT’s large-scale pre-training and sophisticated generation mechanisms contribute to these linguistic characteristics. It provides new empirical evidence for identifying the linguistic features of AI-generated texts and offers practical implications for the teaching of second language writing.