PlainAI AI explained plainly

Khan Academy · 2026 · 04 · 15 · Impact · ~2 min read

AI tutor data came in — Khanmigo's first real-world results

After two years of pilots, Khan Academy's AI tutor Khanmigo reached 1.5 million students in 130 countries. The 2025-2026 randomised trial data showed a measurable but modest jump in maths scores — 12% improvement after a full year — and saved teachers about 37 minutes a week on admin. Real, useful, not magic.

What's actually new

  • A real randomised trial. WestEd ran a controlled study across 47 schools — 0.15 standard deviation improvement on algebra readiness after one semester. Modest but statistically significant.
  • Year-long impact in disadvantaged schools. 12% score improvement in maths and 6-percentage-point jump in independent problem-solving in target schools.
  • Teachers got time back. ~37 minutes per week on admin work, freed up by Khanmigo's progress dashboards and lesson-plan helpers.
  • 1.5 million students, 130 countries by end of 2025 — one of the first AI-in-education products with real-world scale.

If you want more

Worth knowing~30s
  • Writing didn't improve. Maths gains were real; essay-coaching trials showed no meaningful improvement in writing skills. AI tutors help with structured problems, struggle with creative ones.
  • Engagement is uneven. Some students love it, others don't initiate contact with the AI tutor. The product still has an 'open the conversation' problem.
  • Khanmigo schools chose to participate. Selection bias may inflate results vs a random national rollout.
Who should care~20s

Teachers and school administrators considering AI tutors. Parents of students using or considering Khanmigo. Education researchers. Anyone debating whether 'AI in schools' helps or hurts. Policymakers funding ed-tech.

What to do about it~20s

If you're a teacher, the time-savings on admin work are real — try it for that alone. If you're a parent, treat Khanmigo as an extra resource, not a replacement for human teaching or for talking to the teacher. The 12% maths jump matters; the writing zero is honest data — pay attention to what AI tutors don't do.

Honest take~45s

Khanmigo's 2025-2026 data was the most important AI-in-education evidence we got. Most edtech promises don't survive contact with real classrooms; this one mostly did, with caveats. The maths win is real and matters for kids who otherwise wouldn't get one-on-one attention. The writing zero is the honest detail that should temper expectations. Khan Academy proved AI tutors can scale to a million-plus students without breaking; what they haven't proved is that they should be the default, or that all subjects benefit equally. Plain AI's curriculum work sits in the same territory — same realistic expectations should apply.

Other recent impact updates

Sources

Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com