Schoolchildren in a Chinese classroom work on tablets.

China’s AI-Native Generation: Inside the Drive to Teach AI From Childhood

Key Takeaways

  • China is undertaking one of the world's largest AI-assisted learning experiments, covering almost 200 million primary and secondary pupils.
  • By 2030, Beijing wants AI literacy to be universal among schoolchildren, progressing from basic concepts to advanced projects.
  • Nearly half of China's provinces and municipalities have rolled out binding AI education plans, with Shanghai mandating at least 30 hours of AI teaching per grade each year.
  • Some regions, including Ningxia and Gansu, plan to put AI literacy into high-stakes school entrance exams by 2027.
  • The effort is sweeping but uneven, with no agreed definition of AI literacy and no standard way to measure it.
  • Early research is mixed: AI use is linked to higher homework scores but lower exam scores, and heavier use is linked to weaker memory and thinking skills.
  • The drive is explicitly tied to China's technological rivalry with the United States and its ambition to nurture homegrown innovation.

The 13-Year-Old Who Has Built 200 Apps

At a top private school in Beijing, 13-year-old Peng Gaodi has used artificial intelligence to build more than 200 apps, including a homework-submission platform and a secondhand marketplace. He pitches ideas to real investors, idolises Elon Musk, and hopes to become a computer scientist. Around him, classmates program humanoid robots to play soccer, attend coding workshops, and train on flight simulators. Posters across the 30-acre boarding school campus encourage students to "create a virtual world in five minutes with AI vibe coding."

That is the opening scene of a Reuters investigation published on 16 November 2026, "Inside China's drive to build an AI-native generation", based on weeks of school visits, interviews with more than a dozen educators, parents, and experts, and a review of more than 100 academic papers and government procurement records. The school in question is the Beijing New Talent Academy, a 39,000-dollar-a-year private school serving more than 2,000 students, which has built its own digital innovation curriculum covering AI, robotics, and programming, and developed a classroom AI model incorporating features from popular Chinese systems.

Peng's story is the sharp end of something much larger. With almost 200 million students enrolled in primary and secondary education, China is undertaking one of the world's largest AI-assisted learning experiments, with the technology increasingly integrated into teaching nationwide. The ambition is not modest. By 2030, Beijing wants AI literacy to be universal among schoolchildren. Current broad AI education guidelines state that students should progress from understanding basic concepts in their early years to applying advanced tools and critical-thinking skills in complex projects.

The headmaster of the Beijing New Talent Academy, Liu Wei, told Reuters that only a small minority of students at the school would become AI scientists or engineers. For the rest, the objective is to make AI an everyday tool for learning and creating. That distinction matters, because it captures the two different ambitions running through China's push: producing world-class AI talent, and making an entire generation comfortable with AI as a daily instrument.

"In the past, if I couldn't solve a problem at home, I'd have to call my teacher, which increases their workload," Peng told Reuters. "But now I can just ask AI to help me learn better. It noticeably improved my learning efficiency." He added that he tries to solve coding problems himself before asking AI for hints rather than complete answers, and said the technology has helped deepen his understanding of maths. "I think AI is a tool that helps me, not something I depend on."

A State-Led Push Without a Western Equivalent

China views the technology as central to its economic future and technological rivalry with the United States, hoping early exposure will nurture the homegrown innovation needed for major breakthroughs. President Xi Jinping has declared AI education fundamental to reform. That political backing is what turns a set of school experiments into a national project.

The contrast with the West is stark, and Reuters draws it explicitly. Beijing's state-led push far outpaces efforts in the West. By contrast, the United States relies on decentralised, nonbinding guidance and lacks dedicated funding or state-level teacher standards, according to Stanford University's 2026 AI Index. There is no American equivalent of a national plan to make AI literacy universal among schoolchildren by a fixed date.

The coordination is visible in the provincial plans. A Reuters review of provincial "AI+ education" blueprints found a degree of official coordination unseen elsewhere. Nearly half of China's provinces and municipalities have rolled out binding plans. Shanghai mandates at least 30 hours of annual teaching on AI skills per grade. The Beijing city government plans to integrate AI into student health monitoring, as well as set up 100 AI pilot demonstration schools. Regions including Ningxia and Gansu plan to integrate AI literacy directly into high-stakes school entrance exams by 2027.

Since 2024, China's education ministry has encouraged schools to introduce AI courses, train teachers, and develop pilot programmes. Primary-school pupils are introduced to basic AI tools, while older students learn how AI systems work and apply them to increasingly sophisticated projects. The structure is deliberate: exposure early, understanding in the middle years, and application in the later years.

The scale of the spending is large, though hard to pin down precisely. A Reuters review of more than 100 government procurement documents found authorities committed at least 31 million dollars in the past year to AI classroom equipment, training, and teaching platforms. Reuters noted that the true figure is likely far higher, scattered across thousands of local budgets and not officially tallied. Consultancy iResearch estimates 2025 spending on education digitalisation at just under 83 billion dollars, with generative AI projects accounting for 46 percent of AI education procurement. Those figures should be treated with care: the 31 million dollar figure is a floor based on documents Reuters could review, not a total, and the iResearch estimate is a consultancy projection, not an official statistic.

Two Chinas in the Classroom

One of the most striking findings of the Reuters investigation is how uneven the effort is. The same technology is being used in radically different ways in different parts of the country, and the gap between the richest and poorest schools is wide.

At one end is the Beijing New Talent Academy, with its robots, flight simulators, and 39,000-dollar fees. At the other end is Funan County Qinqing High School in rural Anhui province, hundreds of kilometres away, where the school's 4,000 pupils, most of them children of migrant workers, learn core subjects through AI-generated lessons. The school is using the same technology to solve a problem affluent urban areas rarely face: a shortage of qualified teachers.

At Funan, students follow personalised classes on tablets, with AI agents posing questions rather than providing answers when they struggle. "An AI agent doesn't suffer from fatigue, and it never makes a student feel stupid for asking a basic question," the headmaster, Yang Tianliang, told Reuters. The school uses popular Chinese AI models from Alibaba's Qwen, ByteDance, and Z.AI to generate text, images, interactive lessons, and videos. Yang said it had also introduced small-group projects to encourage greater student interaction.

Yang credited the approach with helping lift university admission rates from below 60 percent to 90 percent in recent years, although Reuters could not independently verify the figures. That caveat is important, and this article repeats it rather than presenting the claim as fact. Yang estimated that at least 200 other Chinese schools were experimenting with similar systems.

The rural-urban divide is not only about resources. It is also about purpose. In wealthy urban schools, AI is a tool for creativity and advanced learning. In rural schools, it is a substitute for teachers who are not there. Both are legitimate uses, but they point in different directions, and they raise different questions. A system that uses AI to fill teacher shortages may improve access to education, but it may also entrench a two-tier system in which the poorest children are taught by machines while the richest are taught by people.

An Alibaba spokesperson told Reuters that its open-source AI models and Cloud for Youth Programme give students direct access to AI tools and education. ByteDance and Z.AI did not respond to requests for comment. China's education ministry did not respond to a request for comment.

The Applications Go Further Than Tutoring

The provincial blueprints reviewed by Reuters offer an early look at classroom applications that go well beyond coding and tutoring. Guizhou is integrating AI facial recognition and behavioural analysis into classrooms. Shanghai is deploying systems for mental and physical health monitoring. In Yunnan, 14,000 classrooms are slated to have lessons recorded and analysed by AI, turning Chinese schools into what Reuters called a high-stakes testing ground for the opportunities and risks confronting global education.

These applications raise questions that go beyond education policy. Facial recognition and behavioural analysis in classrooms involve the collection of biometric data from children, on a large scale, with unclear limits on retention and use. Health monitoring systems collect sensitive personal data. Lesson recording creates a permanent record of what happens in a classroom. None of these applications is inherently wrong, and each has a plausible educational justification, but the scale and the lack of public debate about them should give pause. In most Western countries, proposals of this kind would trigger years of argument about privacy, consent, and surveillance. In China, they are being rolled out as part of provincial plans.

It is also worth noting what is not in the reporting. There is no account of whether parents can opt out of facial recognition or health monitoring, or of what happens to the data collected. Those questions were not answered in the material Reuters published, and this article does not attempt to answer them.

The Measurement Problem

For all its scale and ambition, China's AI education drive has a basic problem: nobody has agreed on what success looks like. China has yet to establish a standard framework for measuring AI literacy, leaving schools to experiment and making success difficult to assess. "Mature, widely adopted assessment rubrics for compulsory school-level AI competency are still being built," Selena Guo, portfolio lead at research firm China Policy, told Reuters.

Ted Mo Chen, chief of staff at Chinese education technology startup ClassIn, was blunter. "Implementation is all over the place," he said, adding that few schools he had visited had integrated AI into the curriculum effectively. The Reuters investigation found starkly different approaches across the country, and little consensus on what AI literacy means, how it should be taught, or how it should be measured.

That matters because without a measurement framework, it is impossible to know whether the investment is working. A school can buy AI equipment, run AI lessons, and report that it is teaching AI, without any evidence that students are learning anything useful. The risk is that the drive becomes a procurement exercise, in which the visible signs of AI education, the devices, the platforms, the pilot schools, substitute for the harder work of changing how children learn.

The plan to put AI literacy into high-stakes entrance exams in Ningxia and Gansu by 2027 is one attempt to solve this problem. Exams create a measurable standard, and they force schools to teach to it. But they also create a new risk: that AI literacy becomes another subject to be drilled and memorised, rather than a set of skills to be practised. That tension, between measuring AI literacy and preserving what makes it valuable, is one that every country attempting this will face.

What the Early Evidence Actually Shows

The early research on AI's effects on children's learning offers mixed signals, and Reuters reported it carefully. A recent study examining nearly 27,000 Chinese teenagers found AI adoption was associated with an 18 percent increase in homework scores but a 20 percent decline in monthly examination scores within six months, with the biggest declines in social science subjects. Another nationwide Chinese survey this year, involving more than 320,000 pupils, reported educational benefits from AI but also concerns about dependence and weaker independent thinking. Students using AI at least three times weekly showed lower memory, comprehension, analytical, and creative skills than less frequent users.

The findings do not establish causation, and Reuters was explicit about that. It is possible that students who struggle are more likely to turn to AI, rather than AI causing the struggle. But the pattern reinforces teacher concerns about cognitive outsourcing, where children increasingly delegate mental tasks to technology rather than developing those skills themselves. Chen from ClassIn warned that children's thinking processes could be "completely bypassed" if schools failed to prevent them from relying on AI for quick answers. Chinese academic research reviewed by Reuters also suggested that students in rural areas may be particularly vulnerable to cognitive outsourcing and device addiction.

These findings sit uncomfortably alongside the enthusiasm of the national drive. The same tool that helps a 13-year-old build 200 apps may also be helping other children avoid the hard thinking that learning requires. The difference is not in the tool. It is in how it is used, and in whether anyone is teaching children to use it well.

Some educators and parents are already experimenting with safeguards. At Beijing New Talent Academy, Peng said he tries to solve coding problems himself before asking AI for hints. For Beijing parent Rita Luan, whose 11-year-old daughter attends a leading public school in the capital, the challenge is ensuring that AI strengthens rather than shortcuts learning. Her daughter's school still largely teaches AI concepts rather than practical applications, so Luan supplements lessons at home, teaching her daughter to prompt ByteDance's Doubao chatbot under close supervision. Her rule: spend 15 minutes trying to solve a difficult maths problem before turning to AI. Days later, Luan asks her to solve the same problem without help. "The point is to make sure she actually digested the solution," she said.

That 15-minute rule is a small, practical answer to a very large question. It is also a reminder that the most important part of AI education may not be the technology at all. It may be the habits that surround it.

Why This Matters Beyond China

China's drive matters for reasons that go beyond education. The country views AI as central to its economic future and technological rivalry with the United States, and early exposure is seen as a way to nurture the homegrown innovation needed for major breakthroughs. A generation that grows up building with AI is, in this view, a generation that will invent with it.

That logic is not unique to China. Every country that takes AI seriously is asking the same question: how do we prepare children for a world in which AI is everywhere? The difference is in the answer. China has chosen a state-led, top-down approach, with binding provincial plans, national targets, and exam incentives. The United States has chosen a decentralised approach, with nonbinding guidance and no dedicated funding. Other countries sit somewhere in between. Australia, for example, has published a national framework for generative AI in schools, which we examined in our guide to Australia's generative AI in schools framework, setting out when and how AI may be used in classrooms. The approaches differ, but the underlying question is the same.

For readers outside China, the most useful lesson may be the cautionary one. Scale is not the same as success. A country can put AI in every classroom and still not know whether children are learning. The research findings, the measurement gap, and the uneven implementation are warnings that apply everywhere, not only in China.

For readers inside the AI industry, the drive is also a signal about where Chinese AI is heading. The models being used in classrooms, Alibaba's Qwen, ByteDance's Doubao, and Z.AI, are the same models competing in the wider market. A generation that grows up using them will be comfortable with them, and that comfort is a form of soft power. Apple, for example, has turned to Qwen to power Apple Intelligence in China, as we reported in our piece on Apple Intelligence finally reaching China, powered by Qwen rather than ChatGPT. The classroom is one more place where that ecosystem is taking root. For a broader introduction to the Chinese AI landscape, see our guide to DeepSeek AI, which explains the models, features, and pricing of one of China's most prominent AI companies.

The AI-native generation is not a slogan. It is a policy, with targets, budgets, and exams attached. Whether it produces the innovators China hopes for, or a generation that leans too heavily on machines to think, is a question that will take years to answer. The experiment is already running, and the rest of the world is watching.

What We Could Not Verify

  • The university admission figures cited by the headmaster of Funan County Qinqing High School, a rise from below 60 percent to 90 percent, could not be independently verified by Reuters, and are repeated here only as his claim.
  • The iResearch estimate of 2025 education digitalisation spending at just under 83 billion dollars is a consultancy projection, not an official figure, and the 46 percent share attributed to generative AI projects is an estimate.
  • The 31 million dollar procurement figure is a floor based on documents Reuters reviewed, not a total. The true figure is unknown and likely higher.
  • Whether parents can opt out of facial recognition, behavioural analysis, or health monitoring in schools, and how the data collected is stored and used, was not addressed in the reporting reviewed.
  • The precise content of the national AI education guidelines, including the exact targets for each age group, has not been published in full in the material reviewed.
  • The number of schools experimenting with AI-generated lessons, beyond the estimate of at least 200 given by the Funan headmaster, is not established.
  • The studies cited on homework and exam scores show associations, not causation. The findings should not be read as proof that AI use causes lower exam scores.
  • China's education ministry did not respond to Reuters' request for comment, so there is no official response to the findings in the reporting reviewed.

Sources

Similar Posts