Since the launch of ChatGPT in 2022, generative chatbots have gained popularity in the field of education among both students and educators for many different purposes. To contribute to the understanding of AI-generated texts, this study investigated the differences in linguistic complexity and cohesive indices between AI-generated and human-produced argumentative essays. A corpus of 113 argumentative essays, composed by EFL undergraduate students, were compared to a corpus of 100 argumentative essays generated by ChatGPT. This study found that ChatGPT-generated essays exhibited the linguistic characteristics of the academic genre to a greater extent than essays composed by EFL students. Most importantly, ChatGPT-generated argumentative essays were significantly more diverse and varied than human-produced essays. As for cohesion, only half of the cohesive devices showed greater values for the AI-generated texts. This article concludes with pedagogical implications for educators in the field of applied linguistics.