Subscribe

Industries / Education

AI in Education

AI clearly saves teachers planning time, and well-designed tutors have produced real learning gains in controlled trials. Whether AI tutors work at scale, without weakening pupils’ own thinking, is still unproven and contested.

Profile reviewed 23 Oct 20264 news stories5 timeline entries

cut in weekly lesson-planning time for English teachers using ChatGPT with a guide (EEF trial)
31%
disadvantaged pupils a year England says its planned AI tutoring tools could reach
450,000
of US teachers who say their school has a specific generative AI policy (Harvard survey, 2026)
18%

Updated daily · newest first

Latest in education AI

Policy Brussels publishes two reports on AI in Europe’s classrooms Generative AI is becoming mainstream in EU schools, but the Commission’s own evidence says the effect on long-term skills is still unclear. Applications · Policy

Profile

Reviewed monthly · Sources (12)

Where AI works today

The clearest gains so far are in teacher workload and in carefully designed tutors. In an Education Endowment Foundation trial across 68 English secondary schools, teachers given ChatGPT plus a guide on using it cut weekly lesson-planning time by about 31% (roughly 25 minutes), and an independent panel saw no drop in the quality of their materials. A Harvard randomised trial found students learned more introductory physics from a purpose-built AI tutor than from an active-learning class, in less time. Larger school trials show smaller but real gains: an independent two-year randomised trial of Khan Academy’s Khanmigo in 18 Tennessee middle schools found modest maths gains, similar to those from Khan Academy practice without AI.

Still unproven

Evidence on AI tutors at scale is thin, and results depend heavily on how a tool is designed and used. A PNAS study of nearly 1,000 Turkish high-school maths pupils found unrestricted GPT-4 access raised practice scores, but once access was removed those pupils scored 17% lower on exams than pupils who never had it. An October 2026 randomised trial of a course-integrated AI tutor with 2,379 US undergraduates found lower final grades and much lower course participation, with larger losses for first-generation students. In the Khanmigo trial most pupils rarely used the tutor for real dialogue. England’s own government calls the evidence base for AI tutoring limited. Cheating is the other big worry, and a Harvard survey found only 18% of US teachers said their school had a specific generative AI policy.

Regulation

United States: There is no federal AI-in-education law. A July 2025 Education Department letter confirmed that federal grants can pay for AI tutoring and AI literacy, following an April 2025 executive order. States are moving faster, and New York City has paused student-facing generative AI up to grade 8 for 2026-27. European Union: The AI Act has banned emotion recognition in schools since February 2025. It treats AI used for admissions, grading and exam proctoring as high-risk, and those rules now apply from 2 December 2027. United Kingdom: AI tutoring tools must meet the Department for Education’s generative AI product safety standards, and in March 2026 Ofqual told exam boards to tighten checks on AI misuse in coursework. China: Ministry of Education guidelines from May 2025 bar primary pupils from using open-ended generative tools on their own and stop teachers handing their core role to AI.

Proven deployments

Named organisations, measured results, linked sources only

  1. 31% less planning timeEducation Endowment Foundation (259 teachers, 68 English secondary schools)Teachers who used ChatGPT with a guide spent 56 minutes a week planning Year 7 and 8 lessons, against 82 minutes for the comparison group, with no noticeable drop in quality. Independent trial.Education Endowment Foundation (December 2024)
  2. 0.63 SDHarvard University physics departmentIn a randomised crossover trial, students taught by a purpose-built AI tutor scored higher on post-tests than those in an active-learning class, in less time. Peer-reviewed.Scientific Reports (June 2025)
  3. 0.3 SDWorld Bank with Edo State, NigeriaPupils randomly assigned to six weeks of after-school AI tutoring gained about 0.3 standard deviations. Reported by the programme's evaluators before full publication.World Bank Blogs (January 2025)
  4. 0.14 SDKhan Academy Khanmigo, Hamilton County, TennesseeAn independent two-year cluster randomised trial in 18 middle schools found a full year of active use implied maths gains of about 0.14 standard deviations, similar to practice without AI.NBER Working Paper 35620 (August 2026)
  5. 154 of 156 schoolsEstonia AI Leap20,000 students and 4,800 teachers in 154 of Estonia's 156 upper-secondary schools took part in the 2025-26 pilot. An adoption figure, not a learning result. Self-reported.AI Leap factsheet, e-Estonia (April 2026)

Who’s putting AI to work

  • Khan AcademyNon-profit edtech · Khanmigo tutor, tested in a large independent trial
  • OpenAIAI lab · ChatGPT Edu on campuses, partner to Estonia's AI Leap
  • GoogleAI lab · Gemini for universities, Estonia partner
  • AnthropicAI lab · Claude for Education at universities
  • PearsonPublisher and exam board · supplier in England's AI tutoring pilot
  • Education Endowment FoundationEvidence charity · runs the school AI trials England relies on
  • OfqualExams regulator, England · pressing exam boards on AI misuse in coursework
  • New York City Public SchoolsLargest US school system · one-year student AI moratorium up to grade 8

What to watch

  1. Later in 2026: Ofqual’s decision on on-screen exams, with coursework changes possible after AI misuse concerns
  2. During 2026-27: New York City reviews its student AI moratorium and recommends next steps
  3. 2026-27: Estonia’s AI Leap expands to vocational schools and a new grade 10 group
  4. March 2027: England’s AI tutoring pilot ends, with findings due
  5. End of 2027: England aims to offer AI tutoring tools to schools nationally
  6. 2 December 2027: EU AI Act high-risk rules start for AI used in admissions, grading and proctoring
Sources
  1. 450,000 disadvantaged pupils could benefit from AI tutoring tools (Department for Education, January 2026)
  2. Teachers using ChatGPT with a guide cut planning time by over 30 per cent (Education Endowment Foundation, December 2024)
  3. AI tutoring outperforms in-class active learning (Scientific Reports, June 2025)
  4. Generative AI without guardrails can harm learning (PNAS, July 2025)
  5. AI tutoring with Khanmigo in a two-year school experiment (NBER, August 2026)
  6. Course-integrated AI tutoring: a randomised university trial (EdWorkingPapers, October 2026)
  7. AI and coursework integrity briefing (Ofqual, March 2026)
  8. Guidance on AI use in schools (US Department of Education, July 2025)
  9. China issues guidelines on AI education in schools (Henan Provincial Government, May 2025)
  10. The EU digital omnibus on AI (Taylor Wessing, May 2026)
  11. The AI cheating crisis in schools (Harvard Graduate School of Education, October 2026)
  12. AI Leap factsheet (e-Estonia, April 2026)

Keep exploring

SectionResearchLandmark AI papers in reading paths for students. SectionGlossaryPlain-English definitions of AI terms. CountryAI in the United KingdomEngland's AI tutoring plans and exam rules. CountryAI in ChinaNational AI education guidelines. CountryAI in the European UnionHow the AI Act treats education.

Education on the timeline

Full timeline →
  1. China issues national guidelines on AI in schools
  2. Harvard trial finds AI tutor beats active-learning class in physics
  3. England announces AI tutoring tools for disadvantaged pupils
  4. Ofqual tells exam boards to act on AI misuse in coursework
  5. New York City pauses student-facing generative AI up to grade 8

More industries

All industries →
Healthcare Finance & banking Manufacturing Retail Logistics Legal Media & marketing Defence