evidoria

← Back to browse

Good practice Imported

Medly AI's Micro-Randomised GCSE Science Trial — Higher Attainment After Four Weeks of AI Tutoring

United Kingdom · London · See the United Kingdom profile · See the London profile

Evidence: Randomised controlled trial Top 25% 60/100 · Ask Evidence Copilot about this practice

A four-week, individually randomised trial of 929 English GCSE Biology, Chemistry and Physics students found those using the Medly AI tutor scored significantly higher than peers doing self-directed revision (Hedges' g=0.33), with gains in all three sciences.

0.33 Hedges' g
Effect size, overall attainment
0.52 Hedges' g
Effect size, Biology
0.32 Hedges' g
Effect size, Chemistry
0.31 Hedges' g
Effect size, Physics
30.7%
Attrition rate
Medly AI's Micro-Randomised GCSE Science Trial — Higher Attainment After Four Weeks of AI Tutoring

Details

Maturity
Scaling
Promoter
Medly AI / University College London
Period
2026 (four-week trial)
Keywords
edtech, AI tutoring, secondary education, GCSE science

Context

Medly AI, founded by former NHS doctors and UCL graduates, ran a four-week multisite individually randomised controlled trial of its AI tutoring platform during the 2026 GCSE revision period.

Activities

929 GCSE Biology/Chemistry/Physics students completed a baseline assessment and were randomly allocated to use Medly or continue usual self-directed revision; 644 completed the post-test.

Results

Students using Medly showed significantly higher post-test attainment (Hedges' g=0.33, 95% CI 0.18-0.48), with gains in Biology (g=0.52), Chemistry (g=0.32) and Physics (g=0.31). Positive effects held for both Pupil Premium and non-Pupil Premium students.

Conclusions

Attrition was 30.7% and outcome measures were curriculum-aligned rather than standardised; authors call the results preliminary. Medly has since been used by over 400,000 people in the UK and raised an $8m seed round.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Medly AI research team, participating school teachers

Conditions for success

  • baseline and post-test assessment across multiple sites
  • monitoring attrition and subgroup (disadvantage) effects

Common failure modes

  • 30.7% attrition between baseline and post-test

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful