China Is Running a National AI-Tutor Experiment on 200 Million Kids. Should You Care?

The Experiment Is Already Running

China’s Squirrel AI platform has now reached roughly 52 million students across tens of thousands of schools and learning centers. A peer-reviewed study conducted with SRI International found that students on the platform made significantly greater math gains than peers in traditional classrooms. Those are real numbers from a credible research organization, not marketing copy.

The platform is preparing to enter the US market.

So: should you care?

The short answer is yes, but probably not for the reason you are expecting.

The Part Most Coverage Skips

Every headline focuses on the scale. Fifty-two million kids. Thousands of schools. Measurable gains. What the headlines tend to skip is a quieter detail: the system decides what each student learns, in what order, and at what pace. The student has no input.

That is not a small footnote. That is the whole design.

Chris Dede, a professor at Harvard who studies educational technology, draws a precise line between two things that get conflated constantly. Adaptive learning means the system understands what a student knows and fills the gaps. Personalized learning means the system accounts for what the student wants to know, how they prefer to work, and what they are trying to build toward. Squirrel AI does the first. It does not do the second.

A system that closes knowledge gaps efficiently is genuinely useful. A system that never asks what the kid cares about, and never trains the kid to direct a tool, is a different thing entirely. Both can produce test score gains. Only one produces a kid who knows how to think.

Two Separate Questions for US Parents

When you hear about this experiment, there are actually two concerns worth separating.

The geopolitical one. Tens of millions of kids in another country have had AI-assisted learning as a normal part of school for years. US classrooms, with rare exceptions, are still running on structures that would be recognizable to anyone who graduated in the 1990s. That gap is real.

The more immediate one. When AI tutoring does reach your child’s school or home, and the timeline suggests it will, what position will your kid be in? Will they be a passive subject of the system, following whatever the algorithm serves next? Or will they be someone who understands how these tools work, knows when to question the output, and can frame a problem before handing it to a model?

The second question is the one you can actually do something about right now.

The Driver’s Seat vs. the Back Seat

Here is the thing about AI tutoring at scale: the tool is not the problem. The relationship to the tool is the problem.

A kid who understands how an adaptive learning system works, what data it is using, what it is optimizing for, and what it cannot do, is in a fundamentally different position than a kid who just opens the app and follows prompts. Same tool. Completely different outcome.

This is not a hypothetical distinction. It plays out with every technology kids interact with, from search engines to social feeds to AI writing assistants. The kids who do well with these tools are not the ones who use them the most. They are the ones who understand what the tool is actually doing and can decide when to trust it, when to push back, and when to set it aside.

That is a skill. It is learnable. It is also almost entirely absent from standard school curricula, which tend to treat technology as a delivery mechanism rather than a subject worth understanding.

What Schools Are Not Doing

US schools are not set up to teach this. That is not a criticism of teachers, most of whom are doing serious work under serious constraints. It is a structural observation. A system built around standardized testing and coverage of required content does not have natural room for teaching kids to interrogate the tools they use, run real projects with actual stakes, or develop judgment that transfers beyond the test.

AI tutoring arriving in that environment will not automatically fix the gap. It may make it wider, if kids get better at following algorithmic prompts without ever learning to direct them.

What Actually Helps

The answer is not to keep kids away from AI tools. That ship has sailed, and avoidance is not a skill.

The answer is to make sure that before, during, and after your kid uses any learning technology, they are the one in charge of it. That means:

• Real projects over passive consumption. A kid who uses an AI tool to build something, write something, solve something with actual stakes, learns more about the tool’s limits than a kid who just receives content from it.
• Judgment over recall. Knowing how to evaluate an AI output, spot a gap in its reasoning, or recognize when it has hallucinated something, is more durable than any fact the tool could deliver.
• The kid directs the tool, not the reverse. This sounds simple. It is not the default when you hand a child a screen with an algorithm behind it.

These are not abstract future-of-work concerns. They are practical skills that change how a kid performs right now, in school and out of it.

Where Globeskool Fits

Globeskool is an online school for kids aged 8 to 16 that covers core subjects alongside the skills formal education tends to treat as optional: critical thinking, problem solving, communication, creativity, and working with technology. Not technology as a subject to memorize, but technology as a domain where kids practice having real judgment.

The goal is simple. When AI tutoring tools reach your child’s school, and they will, your kid should already be the person who knows how to run the tool. Not the other way around.

If you want to see where your child stands right now, the free assessment takes about 5 minutes.

Take the free assessment