AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting

There is some evidence to support the idea. However, "learning" studies can be a little disappointing since they usually take place over a short period of time. This one had a week-long timescale: https://www.nature.com/articles/s41598-025-97652-6 I think there's a bigger question that would be hard to answer: what does a consistent long term diet of videos and quick answers do to a person's mind/understanding of the world? Also, a lot of the published research is a year delayed - most I've read recently are using ChatGPT 4 or 4o. When it comes to giving high quality feedback, I think there was an inflection point around Opus 4.6.

The textbook is there so that the model has something reasonably correct to work with and doesn't make up too much wrong stuff to teach. For tutoring, I think the approach in https://www.nature.com/articles/s41598-025-97652-6 is promising. (Prompts are included in the supplementary material on the last page: https://static-content.springer.com/esm/art%3A10.1038%2Fs415... ) They start with an existing collection of worked exercises, give the chatbot access to the full solution and then let students interact with it to get a walkthrough or just check their own solution, depending on how much help the student thinks they need.

https://www.nature.com/articles/s41598-025-97652-6 This isn't study mode, it's a different AI tutor, but: "The median learning gains for students, relative to the pre-test baseline (M = 2.75, N = 316), in the AI-tutored group were over double those for students in the in-class active learning group."