TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Google has published findings from two education studies: a 10-week UK trial involving more than 1,100 students and a yearlong study across 31 Italian schools. The company reports that Gemini-supported tutoring performed similarly to human tutoring on some classroom measures, while Italian teachers using Gemini reported more collaborative lessons. The UK trial did not find gains on STAR standardized test scores during the study period.
Google published findings on October 29 from two studies of Gemini in schools, reporting that AI-assisted math tutoring performed similarly to human tutoring on some measures in a UK trial, while teachers in an Italian study said Gemini-supported planning helped them run more collaborative lessons. The results offer evidence about classroom use, but the UK trial found no improvement in STAR standardized test scores over its 10-week period.
The UK study was a 10-week randomized controlled trial involving more than 1,100 students aged 12 to 15. It ran through Eedi’s school platform and compared static hints, human tutoring and interactive AI tutoring using Gemini capabilities alongside Eedi’s educational resources. Teachers monitored students, tutors provided one-to-one support, and human experts reviewed every AI response before it reached a student.
Google said students receiving AI tutoring performed similarly to students paired with human tutors on platform tasks and intermediate quizzes. Expert reviewers approved 95% of the AI responses, up from 76% in the previous study, which Google said represented an improvement of nearly 25% in pedagogical quality. The company also reported that Gemini guided students rather than giving answers in nearly 99% of sessions, and students worked through the math themselves in more than 85% of chats.
The Italian study covered 31 schools and 1,600 teachers in the Salesians of Don Bosco network, representing more than 24,500 students aged 13 to 18. Over a year, researchers collected classroom activities and teacher self-reports about planning with and without Gemini for Education. Google reported that teachers described feeling more creative in planning and teaching in 79% of cases when using Gemini, compared with 30% without it. The company also reported a 38% drop in traditional front-of-class lectures and nearly three times as much student-led collaborative learning in lessons co-designed with Gemini.
Classroom Gains, Test Score Limits
The findings address two practical questions for schools: whether AI can provide useful individual support at scale, and whether it can help teachers make room for more active classroom work. In the UK trial, the reported similarity to human tutoring on platform tasks and quizzes suggests Gemini-assisted support may be useful when teachers and tutors remain involved. It does not establish that AI tutoring can replace educators, since human oversight was built into the trial.
The results also show why classroom performance and standardized test results should not be treated as interchangeable. Google said students receiving AI tutoring made gains on daily tasks and intermediate quizzes, but those gains did not appear on STAR test scores during the same period. That distinction matters to schools evaluating AI: an approach may improve students’ work on lessons without producing measurable change on a broad standardized assessment within a short trial.
In Italy, the reported shift from lectures toward workshops, group work and peer discussion could affect how teachers organize lessons. However, the findings are based in part on educator self-reports and classroom activity data presented by Google. They do not, on their own, establish that Gemini caused lasting improvements in student achievement or wellbeing.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
DJI Mic 3 Wireless Microphone System (2 Transmitters, 1 Receiver, Charging Case)
As an affiliate, we earn on qualifying purchases.
Two Studies, Different Measures
Google said the UK trial followed an earlier exploratory study with education platform Eedi, which tested an AI tutor designed to identify students’ mistakes and guide them toward correcting their own work. The new trial expanded the evaluation to compare AI support with both static hints and human tutoring, and tracked progress through platform assignments, quizzes and an independent standardized test.
The UK trial’s STAR results require particular care. Google attributed the absence of measured gains partly to the time required for step-by-step tutoring: students averaged about 14 problems per session with AI, compared with 19 when they received static hints. The company said the lower problem volume meant students covered less curriculum during the 10-week study, even as their pacing and understanding improved. This is Google’s explanation of the result, not proof that the shorter coverage was the sole cause.
The Italian research was broader in duration and focused on classroom dynamics and teacher experience rather than the same student test outcomes measured in the UK. The two studies therefore provide different kinds of evidence and should not be read as a single test of Gemini’s effect on learning.
“Technology best serves education when it creates space for inspiring teaching and deeper learning.”
— Google’s report
Questions About Long-Term Outcomes
The UK study did not show gains on STAR standardized test scores over 10 weeks, and the available report does not establish whether a longer period of AI tutoring would change that outcome. Google said it is exploring follow-up trials across a full academic year. It also plans to broaden the measures used to assess what contributes to meaningful learning and later results.
For the Italian study, the published summary does not provide enough detail to independently assess how classroom observations were collected, how comparisons were calculated, or whether reported changes persisted after the study. Google describes the findings as evidence from schools in its study, but the information provided does not establish that the results will generalize to other school systems or settings. The studies also do not settle how AI tutoring compares with teachers across subjects, age groups or longer periods.
Longer Trials and Broader Measures
Google said it is examining follow-up UK trials lasting a full academic year to test whether longer exposure changes learning outcomes and addresses the limits of the 10-week study. The company also plans to expand its measures beyond immediate classroom tasks and quizzes to better track how learning develops over time.
For schools considering these findings, the next evidence to watch includes longer-term standardized outcomes, clearer reporting on research methods and results in settings beyond the participating UK and Italian schools. Until then, the published studies support a measured conclusion: Gemini showed promise in specific classroom uses under educator oversight, while its effects on broader and lasting academic outcomes remain unsettled.
Key Questions
Did Gemini tutoring improve students’ standardized test scores?
Not in the UK trial’s STAR results. Google said students receiving AI tutoring improved on daily platform tasks and intermediate quizzes, but those gains did not appear on STAR standardized test scores over the 10-week study.
How did AI tutoring compare with human tutoring?
Google reported that students using Gemini-supported tutoring performed similarly to students paired with human tutors on platform tasks and intermediate quizzes. The trial included teacher monitoring and real-time review of AI responses by human experts, so it did not test an unsupervised replacement for teachers.
What did the Italian study find about classroom teaching?
Google reported that teachers co-designing lessons with Gemini described greater creativity, while traditional lectures fell and student-led collaboration increased. The study covered 31 schools over one year; the published summary does not establish whether these reported changes led to lasting academic gains.
Why did the UK trial show no STAR score gains?
Google suggested that step-by-step AI tutoring took longer per question, so students solved about 14 problems per session, compared with 19 using static hints. The company said students consequently covered less curriculum during the trial, but the study summary does not prove this was the only reason for the STAR result.
Source: blog.google
