Google brings Guided Learning into Classroom Google brings Guided Learning into Classroom

Two products, different schedules

Google plans a Search pilot that recognizes certain homework requests and offers guided steps. Separately, access to Gemini, Gemini Notebook, Gemini in Classroom and Search will default to off for K–12 student accounts until administrators enable it. Google plans to give schools more details before those control changes arrive.

A short study measured a gain

Google tested Guided Learning with more than 300 University of Tokyo students in a randomized statistics study. Students with standard web tools improved by 5.3 percentage points between tests. Those given Guided Learning improved by 8.1 points, Google reports. The intervention lasted half a day. Participants requested quizzes and visual aids.

Teachers shaped the controls

A separate four-week pilot with Ark covered three UK secondary schools and four subjects. Teachers asked for clearer insight into student activity and tighter focus on assigned material. Google says that feedback led to educator insights and stronger grounding in uploaded curriculum. The resulting Classroom workflow lets a teacher set the learning activity instead of leaving students to start an unrestricted conversation.

Longer trials measured something different

Google published a separate ten-week UK trial on October 29, involving more than 1,100 students using Eedi’s mathematics platform. Incorrect answers triggered a static hint, human tutoring or AI tutoring built with Gemini. Human experts reviewed every AI response before students saw it. Google reported similar performance for AI and human tutoring on platform exercises and intermediate quizzes. Standardized STAR scores did not show the same gains. Students with AI tutoring completed about 14 problems per session, compared with 19 for static hints. Google attributed the difference to the time spent working through each problem. That study makes the measurement choice visible: completing more material, answering immediate questions and improving a broad test score are distinct outcomes. The supervised design also puts the review process inside the intervention being evaluated.

The research has an earlier baseline

In November 2025, Google described an exploratory Eedi trial with 165 UK students aged 13 to 15. It used LearnLM, then incorporated into Gemini 2.5, in chat-based mathematics tutoring under teacher supervision. Google reported a 5.5-percentage-point advantage on novel problems for students receiving short AI tutoring sessions compared with those working with human tutors alone. It also announced plans for further randomized trials across several countries. The new Classroom release follows that progression from bounded tutoring studies to a product teachers can assign. For education software builders, the work unit now includes lesson material, the tutoring conversation and information returned to the teacher.