OpenAI has published findings from a randomized study involving more than 1,000 university students, looking at how ChatGPT and critical-thinking training affect real student work. The study measured performance on an actual course assignment, with attention to answer quality and originality.
The results indicate that students who received critical-thinking training alongside ChatGPT access produced better answers than those who used the model without such preparation. The reported gains include broader thinking and improved originality, suggesting that prompting alone is not enough—students benefit from explicitly practicing how to question, evaluate, and refine AI-generated output.
Because this is a single source, there are no opposing findings to compare. The reported improvement is tied to the combination of ChatGPT and critical-thinking instruction, not to the model used in isolation.