TL;DR
Get travel and outdoor gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Google reported findings from a 10-week UK trial involving more than 1,100 students and a yearlong Italian study across 31 schools. In the UK, students using AI tutoring performed similarly to students working with human tutors on platform tasks and quizzes, but the trial found no gains on STAR standardized test scores. In Italy, teachers reported more creative planning and less lecture-based teaching when using Gemini.
Google has published findings from studies in UK and Italian schools examining how Gemini-supported tools affect math learning and classroom teaching. The UK randomized trial found students using AI tutoring performed similarly to peers receiving human tutoring on platform tasks and intermediate quizzes, while a yearlong Italian study found teachers using Gemini reported more creative planning and fewer lecture-based lessons.
The UK study, conducted with education platform Eedi, involved more than 1,100 students aged 12 to 15 in a 10-week randomized controlled trial. When students answered math questions incorrectly, they received either a static hint, help from a human tutor or interactive AI tutoring that combined Gemini capabilities with Eedi’s teaching resources. Teachers monitored progress and supported students in class; human experts reviewed every AI response before it reached a student.
Google said expert reviewers approved 95% of AI responses, up from 76% in the previous year’s study, describing this as an almost 25% improvement in pedagogical quality. Students receiving AI tutoring performed similarly to those paired with human tutors on tasks within Eedi and intermediate quizzes. According to the report, Gemini guided students rather than directly giving answers in nearly 99% of sessions, and students worked through the math themselves in more than 85% of their chats.
The Italian study covered 31 schools in the Salesians of Don Bosco network and included about 1,600 teachers and more than 24,500 students aged 13 to 18. Teachers reported feeling more creative when co-designing lessons with Gemini: 79% reported that with Gemini, compared with 30% without it. Google also reported that traditional lectures fell by 38%, while workshops, group activities and peer discussion contributed to nearly triple the amount of student-led collaborative learning.
What the Classroom Results Show
The findings offer evidence about two distinct uses of generative AI in schools: guided tutoring for individual students and support for teachers planning lessons. Similar performance to human tutoring on in-platform tasks and quizzes suggests AI could provide another form of help when students get stuck. The trial’s human review and classroom supervision, however, were part of the setup; the results do not establish that unsupervised AI tutoring would produce the same outcomes.
For teachers, the Italian findings point to a possible change in how class time is organized. If planning tools help educators prepare group work and discussion, students may spend more time participating rather than listening to lectures. These reported changes matter to schools weighing AI adoption, but they are not proof that Gemini alone caused improved learning or teacher wellbeing. The source describes teacher reports and classroom activities, not a controlled comparison establishing causation.
The UK result also places a limit on the positive findings: daily tasks and quizzes improved, but standardized STAR scores did not during the 10-week period. Schools and families should distinguish between short-term performance on practiced material and broader achievement measured across a wider curriculum.
Products Worth Considering
DEWALT 20V MAX Cordless Brushless 3-Speed Oscillating Tool Kit (DCS356SD1)
EGO Power+ PST3041 3000W Nexus Portable Power Station with Four 5.0Ah Batteries
EGO Power+ 9-Gallon Cordless Wet/Dry Shop Vacuum (WDV0904) with 56V 5.0Ah Battery and Charger
As an affiliate, we earn on qualifying purchases.
How the Two Studies Were Run
The UK trial followed an earlier exploratory study that tested an AI tutor using Gemini capabilities with Eedi’s platform. In the newer trial, researchers compared three kinds of help after an incorrect answer and tracked students through platform assignments, intermediate quizzes and independent STAR standardized tests. Google said the project included active teacher involvement and real-time expert review of AI responses.
The report offers a possible explanation for why the UK study’s quiz gains did not appear on STAR scores. Students using step-by-step AI tutoring solved about 14 problems per session, on average, compared with 19 for students using static hints. Google said that lower number meant students covered less curriculum during the trial, even as their pace and understanding improved. The company said the trial may not have lasted long enough to cover the range of topics assessed by STAR.
In Italy, researchers worked with the Salesians of Don Bosco school network and the Pontifical Salesian University in Rome over a year. The study gathered classroom activities and teacher self-reports about lesson planning with and without Gemini for Education. Its scale and duration differ from the UK randomized trial, so the results should not be treated as a direct comparison between the two countries.
“Students receiving AI tutoring performed similarly to those paired with human tutors.”
— Google’s report on the UK trial
Limits of the Classroom Evidence
The studies do not establish whether the reported results will hold across other schools, age groups, subjects or ways of using Gemini. Google’s UK trial found no STAR score gains over 10 weeks; the company’s explanation that students covered less curriculum is its interpretation, and the report says further research is needed to examine the pattern.
The Italian results include teacher self-reports and recorded classroom activities. The published material does not establish that Gemini alone caused the reported shifts in teaching, nor does the supplied report excerpt provide enough detail to assess all student engagement or teacher wellbeing measures. It also does not provide a full account of how findings varied across schools or teaching contexts.
In the UK trial, every AI response was reviewed by a human expert before students saw it. The findings therefore do not show how the system performs without that safeguard, or how schools could maintain comparable oversight at larger scale. Google’s article is a company report of its research; independent replication would help clarify how broadly the results apply.
Longer Trials Are Planned
Google said it is exploring follow-up UK trials across a full academic year and plans to expand the measures used to assess meaningful learning and longer-term outcomes. A longer study could test whether students cover more of the curriculum and whether gains on classroom assignments eventually appear on standardized assessments.
The company’s report also directs readers to a technical report with further details on the UK trial. No specific schedule for the follow-up research is given in the material provided. Until more results are available, the current findings describe performance in these particular studies rather than a general guarantee of learning gains from Gemini in classrooms.
Key Questions
Did Gemini tutoring outperform human tutors?
Google reported that students receiving AI tutoring performed similarly to students paired with human tutors on Eedi platform tasks and intermediate quizzes. The report does not say that AI outperformed human tutors, and STAR standardized test scores showed no gains during the trial.
Did the UK study find higher standardized test scores?
No. Google said the gains seen on daily classroom tasks and intermediate quizzes did not appear on STAR standardized test scores during the 10-week trial. The company said students using step-by-step tutoring solved fewer problems per session and may have covered less of the assessed curriculum.
How was AI tutoring supervised?
Teachers monitored students and provided classroom support. In addition, human experts reviewed every AI response before it was shown to a student, so the trial did not test unsupervised AI tutoring.
What did teachers in Italy report?
In the yearlong study across 31 schools, teachers reported greater creativity in lesson planning and teaching when using Gemini to co-design lessons. Google also reported fewer traditional lectures and nearly triple the student-led collaborative learning, based on the study’s classroom measures.
Source: blog.google
