evidoria

← Back to browse

Good practice Imported

EEF/NFER Randomised Trial: ChatGPT Cuts Science Teachers' Lesson-Prep Time by 31%

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

Evidence: Randomised controlled trial Top 56% 47/100 · Ask Evidence Copilot about this practice

A school-randomised controlled trial of 259 Year 7–8 science teachers in 68 English schools found ChatGPT use cut weekly lesson-prep time by 25.3 minutes (31%), with an independent blinded panel finding no drop in resource quality.

56.2 minutes/week
Average weekly lesson-planning time, ChatGPT (treatment) group (summer term 2024 (10 weeks))
81.5 minutes/week
Average weekly lesson-planning time, control group (summer term 2024 (10 weeks))
31 %
Reduction in weekly lesson-planning time (summer term 2024)
EEF/NFER Randomised Trial: ChatGPT Cuts Science Teachers' Lesson-Prep Time by 31%

Details

Maturity
Pilot
Promoter
Education Endowment Foundation & National Foundation for Educational Research
Period
2024
Keywords
generative AI, teacher workload, lesson planning, secondary science education

Context

The Education Endowment Foundation and NFER ran a randomised controlled trial examining whether access to ChatGPT could reduce the time Year 7-8 science teachers spend on lesson preparation, across 68 English secondary schools during the 2024 summer term.

Objectives

To measure whether providing ChatGPT, with a guide for effective use, to science teachers reduces weekly lesson-preparation time without reducing resource quality, compared with teachers using standard preparation methods.

Activities

259 Year 7/8 science teachers were randomly assigned over 10 weeks to a treatment group (ChatGPT access plus an online usage guide) or a control group (standard lesson-prep methods); teachers mainly used ChatGPT to generate questions/quizzes and find activity ideas, and a blinded expert panel assessed the quality of the resulting lesson resources.

Results

Treatment-group teachers spent an average of 56.2 minutes/week on planning versus 81.5 minutes/week for controls, a reduction of 25.3 minutes (31%); the blinded panel found no evidence that resource quality differed between groups.

Conclusions

ChatGPT use, combined with guidance, can meaningfully reduce reported lesson-planning time without a detected quality trade-off, but the trial measured teacher-reported time and resource quality rather than student learning outcomes, and covered only science teaching at one secondary stage with partial-timetable use.

Implementation
Indicative cost
Low (< €50k) — Cost primarily consisted of ChatGPT access and an online usage guide provided to treatment-group teachers, plus the research/evaluation costs of the RCT itself; publication reports no separate cost breakdown.
Time to results
Short (< 1 year) — 10-week trial during the 2024 summer term; full evaluation report published 12 December 2024.
Staffing & skills
259 participating Year 7-8 science teachers across 68 schools, NFER research/evaluation team, blinded expert panel for resource-quality assessment

Conditions for success

  • Providing teachers an online guide for effective ChatGPT use alongside access
  • Random assignment to enable causal comparison
  • Blinded quality assessment of resulting resources

Common failure modes

  • Teachers applied ChatGPT to only part of their timetable, not full teaching loads
  • Measured time/quality perceptions, not student learning outcomes
  • Findings limited to secondary science, one stage of schooling

Where it fits

Governance type
national education research/evaluation body (EEF/NFER)
Scale
multi-school randomised trial (68 schools)
Income level
high-income

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