Middle Schoolers Ignored AI Tutors in Two-Year Study

Students reject AI tutoring at scale

A two-year research study tracking AI tutor adoption across 18 Tennessee middle schools has revealed a fundamental problem with classroom automation: students don't want to use it. Despite near-universal access to Khan Academy's Khanmigo generative AI tutor, middle school students largely ignored the tool during their school day, according to findings detailed in a new research paper.

When students did engage with the chatbot, they rarely used it for its intended educational purpose. Instead, they tested its boundaries with off-topic prompts like "can you write an essay about the history of pizza?" or requests for jokes. Many attempted to manipulate the system into simply providing correct answers rather than working through problems.

"Access was nearly universal but engagement was thin," the researchers wrote. "It may be that human attention is the key ingredient for realizing the full benefits of personalized learning."

Why it matters

School systems and edtech companies are investing heavily in AI tutoring platforms, betting they can address teacher shortages and personalize instruction at scale. This study suggests that technology alone cannot solve educational challenges when student motivation and human guidance are absent. The findings arrive as educators grapple with declining reading comprehension and increased academic dishonesty in the AI era.

Modest gains don't settle the debate

The study did find that students using Khanmigo showed faster math progression compared to a control group without AI access. However, researchers and Khan Academy leadership acknowledged this improvement doesn't validate standalone AI tutoring as an educational solution.

"If a keen, engaged student wants to use the technology as a good tutor, it's ready to go," said coauthor Philip Oreopoulos from the University of Toronto. "The issue is that most students don't use it that way."

Khan Academy CEO Sal Khan told education publication Chalkbeat that the research "put in huge spotlight — not that it's a shock to educators — is how much motivation matters."

Human integration proves essential

The Tennessee findings align with separate research from Stanford published earlier this month, which concluded AI tutoring requires substantial human integration to succeed. Stanford researchers found that AI tools designed to support human tutors produced stronger results than those aimed directly at students.

"The effectiveness of AI-led tutoring depends as much on integration, such as by teachers in classrooms or parents at home, as on the quality of the software itself," the Stanford team determined.

Oreopoulos noted the core tension: "The same technology that can be used as a good tutor, can also be used to make your life easier. But making your life easier equals less effort, less learning."

The research underscores that AI chatbots cannot yet navigate the delicate balance between providing educational support and maintaining the productive struggle necessary for learning. Without active human oversight, students default to the path of least resistance.

These details were first reported by Futurism.