ChatGPT and similar AI tools are rapidly making their way into language classrooms and testing, raising important questions for teachers: Are these texts really suitable for our learners? Do they truly reflect human language use?
This talk draws on a corpus study comparing AI-generated academic texts with human-authored ones to examine how well ChatGPT adapts to different registers (journal articles and textbooks) and disciplines (biology and history). While ChatGPT often produces fluent, coherent writing, our findings show that its texts do not always align with the communicative purpose of the register and often fail to mimic human linguistic choices.
For teachers, this means that ChatGPT can be a helpful tool for generating practice or testing materials, but it should be used carefully and critically. I will share concrete insights from this research, along with survey data from another study that complements the linguistic analysis with human perceptions.
Presenter:
Nur Yağmur Demir is a Ph.D. candidate at Northern Arizona University. She received her BA (Boğaziçi University – 2016) and MA (Middle East Technical University – 2021) in English Language Education in Turkey. She has several years of experience in teaching English and Turkish in EFL contexts. She is interested in register appropriateness and humanlikeness of AI- generated texts, corpus linguistics, English as a Lingua Franca, and computer-mediated communication.
Date & Time:
Wednesday, October 8, 2025, 5:00-6:00 pm PT
Hosted by CIRT-IG:
Elizaveta (Ellie) Kuznetsova & Margi Wald
#catesolcirt