Published on

A $850 Million Experiment That Lasted One Semester β€” Lessons from South Korea's AI Textbooks

"World's first." Few phrases build higher expectations. In March 2025, South Korea introduced AI-powered digital textbooks into public schools β€” the first country in the world to do so. The budget: approximately $850 million. The plan was to start with math, English, and informatics and eventually expand to all subjects across all grade levels. But just one semester later, the program effectively came to a halt. What happened?

Table of Contents

  1. What the AI Textbook Program Promised
  2. The Reality After One Semester
  3. What Teachers Said: The Heart of the Problem
  4. The Smartphone Ban: A Reaction to Digital Overload
  5. What We Can Learn from This

1. What the AI Textbook Program Promised

The central idea behind the AI digital textbook project was straightforward: provide every student with a personalized learning experience. If a teacher can't track the individual level of 30 students simultaneously, what if AI could help? It would identify in real time where a student was stuck and serve up tailored problems and explanations.

When the program launched in March 2025, the target students were in grades 3–4 of elementary school, the first year of middle school, and the first year of high school. The subjects were math, English, informatics, and Korean language for students in special education. The plan was to add one grade per year until it covered all schools and all subjects.

2. The Reality After One Semester

Reality diverged from the vision. Actual classroom adoption in the first semester hovered at around 30%. In practical terms, 7 out of every 10 schools weren't meaningfully using the AI textbooks in their lessons.

By the summer of 2025, the government stripped the AI textbooks of their official "textbook" status and reclassified them as supplementary materials. Government funding was effectively cut. EdTech companies that had developed the content found themselves without a foundation overnight.

What went wrong?

1) Errors and reliability issues: Some AI textbooks contained factual inaccuracies and unclear explanations. Trust began to erode among teachers and parents.

2) Privacy concerns: AI textbooks collected students' learning patterns in real time. But explanations of how this data was stored and who could access it were inadequate.

3) Screen fatigue: Concerns accumulated among teachers and parents about a learning model that required students to stare at digital screens all day.

4) Increased workload for teachers: Using AI textbooks effectively required teachers to learn new platforms, interpret data dashboards, and redesign their lessons. The burden was far greater than anticipated.

3. What Teachers Said: The Heart of the Problem

The most revealing number came from a survey of 2,626 teachers by the Korean Teachers and Education Workers' Union: 98.5% said their prior training was insufficient.

That is nearly everyone. The World Education Blog summarized the core issue: "Moving to AI textbooks requires teachers to become facilitators of learning rather than impart learning themselves, and it also requires teachers to have a strong level of digital skills themselves." The tools arrived before the people who were supposed to use them were ready β€” and without the time or support to build those skills.

This pattern β€” technology leading, people catching up β€” is a recurring mistake in education reform. Even the best tool is useless if the people using it aren't prepared.

4. The Smartphone Ban: A Reaction to Digital Overload

While the AI textbook debate was at its peak, South Korea's National Assembly passed a completely different kind of legislation in August 2025: a nationwide ban on smartphones in K-12 classrooms. The bill passed 115 to 31, with 17 abstentions, and took effect in March 2026.

What drove this? The data was stark. A 2024 government survey found that approximately 43% of South Koreans aged 10–19 were classified as overly dependent on smartphones β€” nearly double the national average.

Exceptions are built in: teachers can allow phones for educational purposes, emergencies are exempt, and students with disabilities can use them as needed. South Korea joined France, which enacted a similar school smartphone ban in 2018, as one of the few countries to enshrine this in law.

The juxtaposition is striking: AI textbooks (more digital) and a smartphone ban (less digital) were both pushed forward in the same period. This tension reflects a society still working out how technology fits into education.

5. What We Can Learn from This

The AI textbook program failed β€” but it teaches.

First, technology without teacher readiness produces backlash. South Korea's ongoing rollout of its 2022 Revised National Curriculum must incorporate the lessons of this experience into how teacher professional development is designed.

Second, student data protection must come first. Before an AI education platform is introduced into public schools, clear legal standards and transparent disclosure about data collection must be in place.

Third, speed matters less than design. Being "world's first" is worth less than building something that actually works. A pilot-validate-scale approach was needed from the start.

South Korea's 2022 Revised National Curriculum is being phased in through 2025–2026, reducing required core subject credits, adding new electives like climate literacy and financial education, and establishing digital literacy as a foundational competency. The direction is right. But the AI textbook episode reinforced that how you get there matters just as much as where you're going.

Do you think the AI textbook project should be revised and relaunched, or rethought from the ground up? We'd especially love to hear from educators in the field.

Further Reading

Sources

Discuss on Twitter β€’ View on GitHub
A $850 Million Experiment That Lasted One Semester β€” Lessons from South Korea's AI Textbooks | MINSSAM.COM