AI-Tutored Students Show Double the Learning Gains

Students learning with artificial intelligence tutors made more than double the progress of those in traditional classrooms, according to a groundbreaking Harvard study. The findings offer hope for addressing declining math and reading scores that have plagued American schools for a decade.

Alpha School, a private education company, is betting big on this technology. The network plans to expand to roughly 50 campuses across the United States this fall, up from just a handful of locations.

Their model is straightforward. Students spend two hours each morning working with an AI tutor that adapts to their individual needs, assesses their progress in real time, and provides immediate feedback. Afternoons focus on life skills like coding, entrepreneurship, and public speaking.

Carl Hendrick, a senior learning scientist at Alpha School, compares training the AI to teaching a self-driving car. Just as autonomous vehicles learn from thousands of miles handling unexpected situations, the educational AI learns from countless student misconceptions and learning challenges.

The Harvard experiment that caught researchers' attention tested an adaptive AI tutoring model in an introductory physics course. Students using the AI showed median learning gains more than twice as high as those using standard active learning strategies.

Kelly Miller, a Harvard professor who co-authored the study, believes adaptive learning represents the future of education. She calls Alpha School's approach "better than what currently exists in K–12."

The Bright Side

The technology arrives at a crucial moment. One in five American students is chronically absent, and academic performance has been sliding nationwide. A 2025 review of 28 studies involving nearly 5,000 students found generally positive effects from AI-driven tutoring systems.

Education researcher Dan Goldhaber from the University of Washington says Alpha School's model resembles other established adaptive learning technologies that have proven effective. The difference is scale and ambition.

Hendrick argues the approach solves a persistent problem in traditional education. "You have massive variance in teacher quality and pupil outcomes in a way that most people would find shocking," he says. The variance between teachers within the same school often exceeds the variance between different schools entirely.

Not all experts are convinced AI should replace human teachers entirely. Greg Kestin, lead author of the Harvard study, worries students might develop "brittle knowledge" that works well on tests but doesn't transfer to real-world situations.

Still, the promise of consistent, personalized instruction at scale offers new possibilities. As Hendrick notes, you can't easily replicate great teachers across thousands of classrooms, but you can scale proven AI models.

The company aims to eventually reach a billion kids with this technology, transforming how students learn across the globe.