How Gemini is helping students and teachers in the classroom

Technology best serves education when it creates space for inspiring teaching and deeper learning. To better understand how AI tools perform in everyday educational settings, we've spent the past year working alongside educators and researchers in classrooms across the UK and Italy.

From testing one-on-one AI math tutoring to analyzing thousands of hours of classroom collaboration, today we're sharing new research findings that demonstrate the impact of this technology on the process of teaching and learning. Together, the results show that AI tutoring using Gemini capabilities can deliver support on par with human tutors for UK students when used responsibly with human oversight, while Gemini itself can foster more collaborative classrooms and support teacher wellbeing in Italy.

To truly understand how student learning is being shaped, we need to study it over time across diverse classroom settings. We previously ran an exploratory study with Eedi, which found that AI tutoring could operate effectively in real classrooms. The study tested an AI tutor combining Gemini capabilities with Eedi's study platform, designed to identify a student's specific mistake and guide them to fix it themselves.

This year, we set out to better understand Gemini's pedagogical capabilities and the nuances of student learning in structured mathematics lessons. We conducted a 10-week randomized controlled trial (RCT) with over 1,100 students (ages 12-15) in the UK to measure their learning gains in math through sessions on Eedi's school platform. Educators were actively involved at every stage of the study - with teachers monitoring student progress and providing hands-on support in the classroom, and tutors guiding students one-on-one through the platform.

Whenever a student answered a question incorrectly, they received one of three types of support: a static hint, help from a human tutor, or interactive AI tutoring combining Eedi's pedagogical resources with Gemini. To ensure teaching quality, human experts reviewed every AI response in real time before it reached a student. While test scores aren't the only measure of learning outcomes, over the course of the trial, we tracked their progress across regular assignments in the Eedi platform, intermediate quizzes, and independent STAR standardized tests.

This trial revealed major leaps in model quality and student performance. Compared to last year's study, Gemini's pedagogical quality improved by nearly 25%, increasing the outright approval rate by expert reviewers from 76% to 95%. Across in-platform tasks and intermediate quizzes, students receiving AI tutoring performed similarly to those paired with human tutors. Just as importantly, AI tutoring promoted active thinking rather than shortcuts: following a pedagogical prompt we provided, Gemini guided students instead of handing them answers in nearly 99% of sessions, and students worked through the math themselves in more than 85% of their chats.