Revising for GCSE science means a lot of biology, chemistry and physics to get through, usually with a textbook and past papers and not much feedback until the exam. Students who get stuck often stay stuck. Teachers can only give so much individual feedback across a full class, especially in the run-up to exams.
The clever bit
Medly is an AI tutor built for exam revision. Researchers from WhatWorked Education and Durham University ran a four-week randomized trial in English secondary schools, assigning some students to revise with Medly and others to revise as they normally would. The trial covered all three sciences, and students' progress was measured with short topic tests marked out of 35 at the end of the four weeks.
Of 929 students who started, 644 completed the final topic tests. Students who used Medly scored about 2.6 marks higher out of 35 than classmates doing their usual revision, an effect size of 0.33 across biology, chemistry and physics. Randomizing students to the tutor or to their usual revision is what lets the researchers compare the two groups directly.
Hooray
A third of a standard deviation in four weeks is an encouraging result for a short stretch of revision with an AI tutor. Short, steady gains like this are what many teachers look for when choosing revision tools, as long as the evidence keeps holding up.
What comes next
Medly funded the study, and the preprint hasn't been peer-reviewed. About 31% of students didn't finish, and the topic tests were marked by AI and checked by teachers. The trial says nothing about final GCSE grades or whether the gains last. Independent replication would help settle how much of the effect comes from the tutor itself.