Khan Academy 'Khanmigo' - Study Finds No Extra Benefits from AI

I've just read this article, "Students didn’t get answers from Khanmigo. They didn’t want its questions, either", which was very interesting. Basically the article is reporting on a large study of Khan Academy's math program that found no extra benefits from AI. At the core of the story is Khan Academy releasing an AI chatbot in 2023. The chatbot would help students by acting like a human tutor, withholding answers, providing hints and guiding students with questions. However, the study that came out in August (2026) suggests that creating an AI tutor is only one hurdle among many in helping students improve their maths skills. 

"Students have to actually want to use it."

Why this article and the results of the study resonated with me, was that in June (2026) I wrote about the same issue in a post called "Underwhelmed by Google Gemini 'Guided Learning' Feature". In this post I wrote about my experience in trying out Google's 'Guided Learning' and noted how easy it actually was for it to spew out the answer and also to sidetrack the chatbot to talk about other things. I've also recently tried out Microsoft's 'Study and Learn' product, which does a similar thing to Google's and Khan Academy's products. Interestingly, when using the Microsoft's product it wasn't so easy to get it to provide me with an answer, but I could sidetrack it, although it still kept asking me questions about the topic that I was sidetracking it with! Based on what the study found, that

"when Khanmigo refused and instead asked questions or offered hints, students largely stopped using it,"

then this Study and Learn activity may well come to an abrupt end with many learners. 

One of the points I made in my own post was this, "After tinkering with Gemini Guided Learning for a couple of hours, it very much appears that the learner needs to be very engaged with the learning to make the most from this feature as it is very easy to bypass any learning if the user wants to." A 15 year old me, who would have just wanted the answer for my homework, would certainly not have been engaged with the learning in the way that these products are designed. The only way I might have learned something is to have a human tutor / teacher helping me through the problem. 

I also wrote that I felt like I was using an EdTech personalised learning program that have been around for many decades. Other than using generative AI chatbot, this approach, question and then based on the users answer another, is very similar. Hence, I'm sure many of the concerns about this form of learning such as the one I mentioned about about sustaining student motivation or engagement (distraction) and self directed study, will be the same issue as it has often been in personalised learning approaches using digital technologies. This study basically found that using the AI product, added nothing to students learning. It will be very interesting to see the results of the UK governments tutor programme when it launches next year.