research review

A review of early ChatGPT research separates model performance from evidence about student learning

Qian Liu, Anjin Hu, Tehmina Gladman, Steve Gallagher

Innovative Higher Education

A scoping review published in 2025 maps 39 studies from ChatGPT’s first eight months in higher education. It finds varied model performance, promising classroom uses, and repeated concerns about accuracy and assessment. Much of the evidence evaluated what the model could produce or what users believed. The review calls for stronger explanations of learning and should be read as an early baseline, not an evaluation of current models.