Why did Sal Khan admit the AI revolution hasn’t happened yet?

In the April 9, 2026 Chalkbeat interview, Khan stated bluntly: “We haven’t seen the revolution I hoped for. The technology is there, but the ecosystem isn’t.” He pointed to three factors: schools lack the hardware to support one-on-one AI tutoring, teachers are not trained to integrate AI into lesson plans, and students often reject AI tutors as “boring” or “creepy.” Khan Academy’s own data shows that only 15% of Khanmigo users engage with the tool for more than 10 minutes per session, suggesting shallow adoption rather than deep learning transformation.

What does Khanmigo’s failure reveal about AI in education?

Khanmigo’s struggles are not about AI capability—GPT-4 can tutor effectively in controlled tests—but about the messiness of real classrooms. Teachers report that Khanmigo often gives overly verbose explanations, confuses students with irrelevant examples, and cannot adapt to the social dynamics of a class of 30 students. A 2025 study by the RAND Corporation found that only 12% of teachers using AI tools felt they improved student outcomes, while 45% cited increased frustration. The real lesson: AI tutoring works in theory but fails in practice because it ignores the human need for trust, brevity, and classroom management.

Who wins and who loses from Khanmigo’s slow adoption?

The biggest winners are narrow-focus edtech companies like Duolingo and Carnegie Learning. Duolingo’s AI language tutor, Duolingo Max, has 2.5 million paid subscribers as of Q1 2026, up 40% year-over-year, because it sticks to a single subject and uses gamification to keep students engaged. Carnegie Learning’s MATHia, an AI math tutor used in 6,000 schools, reports 20% improvement in test scores because it integrates directly into existing curricula. The losers are Khan Academy and any company selling broad “AI tutor for everything” products. School districts are increasingly wary of large, expensive AI platforms; a 2026 survey by EdWeek found that 68% of district leaders prefer subject-specific AI tools over general tutors.

What should Khan Academy do differently?

Khan Academy must pivot from a general AI tutor to a teacher-assistance tool. Instead of replacing teachers, Khanmigo should help teachers with grading, lesson planning, and personalized worksheets—tasks that teachers actually want AI for. A 2026 survey by the Gates Foundation found that 74% of teachers would use AI for administrative tasks, but only 22% want it for direct student tutoring. Khan Academy should also partner with school districts to provide hardware and training, not just software. Without these changes, Khanmigo will remain a niche experiment.

Prediction 1: Duolingo will acquire a math AI startup by Q2 2027 to challenge Carnegie Learning.

Prediction 2: Khan Academy will pivot Khanmigo to teacher-assistance tools by Q1 2027, abandoning the general tutor model.

Prediction 3: The US Department of Education will issue guidance by Q4 2026 that discourages schools from adopting broad AI tutors, favoring subject-specific tools.

• March 2023: Khan Academy launches Khanmigo, a GPT-4-powered AI tutor, with Sal Khan predicting a revolution in education within two years.
• June 2024: Khanmigo reaches 500,000 users, but teacher feedback highlights issues with verbosity and irrelevance.
• January 2025: RAND Corporation study finds only 12% of teachers using AI tools report improved outcomes.
• April 2026: Sal Khan admits in Chalkbeat interview that the AI revolution hasn’t happened, citing ecosystem failures.

• Chart title: AI Tutor Adoption in K-12 Schools (2024-2026, estimated)
• Labels: Khanmigo, Duolingo Max, Carnegie Learning MATHia
• Data (users in millions): Khanmigo: 0.5 (2024), 0.8 (2025), 1.0 (2026); Duolingo Max: 1.0 (2024), 1.8 (2025), 2.5 (2026); MATHia: 0.8 (2024), 1.2 (2025), 1.5 (2026)
• Note: All figures are estimated based on public reports and industry analyst projections.

Article Summary

• Sal Khan’s admission confirms that AI tutors are not ready for broad classroom use due to ecosystem failures, not technical ones.
• Focused AI tools like Duolingo Max and Carnegie Learning MATHia are outperforming general tutors because they solve specific problems.
• Teachers are the gatekeepers of AI adoption—tools that assist teachers, not replace them, will win.
• Khan Academy must pivot to teacher-assistance or risk becoming irrelevant in the edtech market.
• The next 18 months will see a market shakeout, with narrow AI tools dominating and general tutors fading.