Press release: March 2026
Third Space Learning, a leading AI tutoring and education technology provider, is partnering with Stanford University and Cornell University on a Gates Foundation-funded, two-year research and development initiative to advance effective, evidence-based AI math tutoring in US public schools.
The $1.9 million project, led by Stanford University in partnership with Cornell University and Third Space Learning, will fund large-scale research into how spoken, one-to-one AI tutoring can be implemented effectively, responsibly, and equitably in real classroom settings. As part of the partnership, Third Space Learning will receive funding to establish district partnerships and develop the technical infrastructure to support effective experimentation and collaboration between leading researchers, teachers and students at scale.
Grounding AI tutoring in proven high-impact practice
While interest in AI tutoring has accelerated rapidly, Third Space Learning is unique in its focus on building an AI tutor that faithfully reflects the established research behind high-impact, high-dosage tutoring – one of the most effective interventions for closing the math achievement gap.
The project is founded on a shared belief among the partners that AI tutoring will only transform learning outcomes if it is grounded in well-established research into effective tutoring. That work must also be developed in partnership with classroom leaders in real school environments, rather than in isolation from instructional practice.
Third Space Learning brings more than a decade of experience delivering high-impact tutoring, having supported over 170,000 students across more than 4,000 schools and delivered millions of hours of one-to-one math tutoring. This expertise has informed the development of its spoken AI tutor, designed to mirror the dialogue, pacing, questioning, and assessment strategies proven to drive learning gains in human tutoring.
Reaching thousands of students and building the infrastructure to scale
Through partnerships with US school districts, the project will provide weekly spoken AI math tutoring to over 3,000 elementary and middle school students in the second year, with a focus on districts serving historically underserved communities.
Alongside direct student impact, the grant will fund the development of robust learning-engineering infrastructure at Third Space Learning. This will enable rapid, randomized experimentation, rigorous evaluation, and continuous improvement of AI tutoring in real classroom conditions, in partnership with leading research organizations.
The research will examine how technical, programmatic, and student-level factors – such as language model selection, tutor dialogue and pacing, session structure, dosage, and classroom integration – influence learning outcomes, engagement, and equity. Findings, along with de-identified data sets and the learning-engineering infrastructure, will be shared openly with districts, researchers, and the wider education sector to support the responsible scaling of effective AI tutoring and education research efforts nationwide.
Aligning with leading US education and AI researchers
By partnering with Stanford’s SCALE Initiative and the Cornell-led National Tutoring Observatory (NTO), the project aligns Third Space Learning with some of the most influential organizations shaping the future of tutoring and education research in the United States and globally.
“This project represents a major step forward for both Third Space Learning and the field of AI tutoring in education,” said Tom Hooper, Founder and CEO of Third Space Learning. “There is enormous excitement around AI in education, but real impact will only come if AI tutoring is built on the same evidence, discipline, and safeguards as high-impact human tutoring. That focus is why this partnership exists. Working alongside Stanford and Cornell to test and scale spoken AI tutoring in real US classrooms gives us a unique opportunity to gain evidence into the impact of Third Space Learning’s AI tutoring. It is validation of our personalised, adaptive and teacher-led approach to AI tutoring from the start and a critical step toward closing the math achievement gap at scale.”
“AI tutoring should be tested with the same rigor as high-impact human tutoring, and this collaboration is designed to do exactly that,” said Rene Kizilcec, Associate Professor at Cornell University who leads the NTO.
“There is understandable enthusiasm about the potential of AI in education. But if AI tutoring is going to meaningfully improve student outcomes, it must be grounded in what we already know works,” said Susanna Loeb, Professor at Stanford University and faculty director of the SCALE Initiative. “This partnership will rigorously test whether spoken AI tutoring can reflect the core elements that make tutoring effective – strong instructional design, consistent dosage, and alignment with classroom practice.”
The award marks a significant milestone in Third Space Learning’s US expansion and reinforces its position as a leader in research-driven AI tutoring, combining the scalability of AI with the rigor of proven instructional practice.
Download the press release