ChatGPT’s engagement in essay writing has been examined in a new academic study, revealing that while it can produce grammatically accurate and coherent argumentative essays, it struggles with captivating the reader—an essential skill that human student writers excel at.
The research was conducted by Professor Ken Hyland from the University of East Anglia in association with Professor Kevin Jiang of Jilin University. They scrutinised 145 essays created by UK university students alongside another 145 produced by OpenAI’s ChatGPT. The findings, recently published in the journal Written Communication, indicate that student-written essays excelled in employing rhetorical strategies like posing questions, personal insights, and direct appeals—elements that foster a persuasive and engaging bond with the reader.
Since ChatGPT’s public release, it has instilled considerable concern among educators worried about students utilising it for their assignments, according to Professor Hyland. This research aimed to explore how closely AI can replicate the nuances of human essay writing, especially regarding reader engagement.
The phrase “engagement markers” refers to components that actively draw the reader in, such as rhetorical questions, personal comments, or directly addressing the audience with terms like “we” or “you.” These rhetorical elements were predominantly absent from the essays generated by AI, which, despite being structurally sound, appeared more impersonal and detached.
The ChatGPT generated essays imitated academic writing standards but failed to incorporate a personal touch or showcase a clear viewpoint, noted Hyland. They tended to shy away from questions and offered limited personal commentary. In general, they were found to be less engaging, less persuasive, and lacked a strong opinion on various topics.
Conversely, the student essays were documented as being more interactive, employing a broader array of techniques to navigate the reader through complex arguments. This included a markedly greater number of questions and narrative asides—conversational elements that humanise the writing and foster a collective intellectual adventure between the writer and the reader.
The study also provided insight into the mechanics behind ChatGPT’s outputs. Because it operates on statistical modelling from extensive training data, it often produces contextually relevant text but lacks the nuanced engagement characteristic of human writing. This so-called “audience blindness” of the AI—its difficulty in envisioning a specific reader—emerges as a fundamental limitation in crafting persuasive prose.
However, the researchers do not dismiss the role of AI in educational contexts. Instead, they advocate for its thoughtful incorporation within teaching. When students enter educational institutions, they are not merely learning how to write but also how to think—a unique capacity that no algorithm can duplicate, reaffirmed Hyland.