A 2025 survey by the China Center for Youth Research found that over 60% of primary and secondary school students have used AI, with nearly 20% using it frequently. Elementary school students have the highest proportion among all grade levels for using AI to chat and share their inner thoughts.

Author and source: GeekPark

"Artificial rearing"—could it become a privilege in the future?

A child aged four or five is holding a stuffed toy and talking to it.

It’s not about pressing a button to listen to a recorded story—it’s a conversation. The toy calls the child by name, remembers yesterday’s chat, and asks if they had a good day at kindergarten. The child giggles in response, as if sharing secrets with a real friend.

This scene is becoming increasingly common in Chinese households in 2025.

A 2025 survey by the China Center for Youth Research of 8,563 primary and secondary school students found that over 60% have used AI, and nearly 20% use it frequently.

Notably, among all age groups, elementary school students have the highest proportion of using AI for creation, chatting with AI, or sharing their thoughts—not middle or high school students, but the youngest group.

Meanwhile, only 32.1% of households have established rules for their children’s AI usage.

Sixty percent of children are using it, but seventy percent of parents aren't monitoring it.

Ten years ago, people were anxious about children refusing to put down their iPads. Back then, the problems—excessive screen time, declining eyesight, and reduced attention span—were already enough to trouble parents.

Now that AI has arrived, what impact does it have on children's development?

01. Native AI Generation

As of December 2025, the number of generative AI users in China reached 602 million. Among internet users under 19 years old, the usage rate was 29.1%.

For this generation of children, AI has never been a new technology that needs to be "learned how to use"—it’s as natural as running water, available the moment you turn it on.

But children are not just using AI to check their homework.

When breaking down the data from the Youth Research Center, high school students primarily use AI to assist with studying, middle school students enjoy using it for gaming support, while elementary school students have the highest proportion among all age groups using AI for chatting and sharing their feelings. This suggests that for the youngest users, AI is less of a tool and more like a friend always available online.

And this "friend" was not found by the child themselves, but was handed directly to them by the entire supply chain.

In the first half of 2025, AI toy sales on JD.com increased sixfold quarter-over-quarter and grew by more than 200% year-over-year. That same year, over 15 children’s AI hardware companies secured funding, and more than 30 new products were launched. According to data from China’s Ministry of Industry and Information Technology, the Chinese AI toy market reached RMB 29 billion in 2025, with an annual compound growth rate exceeding 50%.

This is a complete chain spanning from large models to chip modules, to toy manufacturing, and finally to live-streaming sales on TikTok. It is highly efficient and extremely fast.

But there is a subtle misalignment in this chain.

Parents born in the 80s and 90s are precisely targeted by the marketing pitch of an “screen-free education miracle”—AI toys that can help children practice English, solve puzzles, and explain science without requiring screen time, sounding absolutely perfect.

But if you examine the actual design of these products, their core selling point is not education, but emotional companionship. Phrases like “a warm, intelligent companion,” “an AI friend who understands emotions,” and “an indispensable partner on your child’s journey of growth” are actual wording used in product launches.

Parents thought they were buying an advanced interactive learning device, but what was actually being sold was a "connection."

This misalignment is the starting point for all subsequent issues.

02. A risk spectrum wider than a smartphone screen

In the latest episode of the Hard Fork podcast, Dana Suskin, a professor at the University of Chicago and a pediatric surgeon, offered a compelling analogy.

She said that many people compare the impact of AI on children to social media, which is understandable but not entirely accurate. A better analogy is processed food.

Whole wheat bread is a processed food, and Doritos are also a processed food—you can’t discuss them at the same risk level. The same goes for AI: asking DouBao, “Why did dinosaurs go extinct?” and discussing the answer with your child is like eating whole wheat bread; having an AI plush bear chat with your child for two hours every day as an emotional outlet is like consuming ultra-processed snacks.

The risk spectrum of AI is much wider than that of a screen. The best use of AI is more educational than the best use of a screen; but the worst use of AI is more dangerous than the worst use of a screen.

The root cause is that traditional screens and AI have entirely different risk mechanisms.

Traditional screens are "passively occupied," while AI is "actively replacing."

Children scrolling through short videos is essentially a one-way consumption of attention. It displaces time that could be spent on outdoor activities and interpersonal interactions, but children understand it’s just “watching something.” AI companionship is different. It remembers your name, picks up conversations from yesterday, and proactively checks in on your emotions—these are precisely the core characteristics of intimate relationships: continuity of memory, personalized interaction, and emotional responsiveness.

Research finds that children with social anxiety, low self-esteem, or a lack of security at home are more likely to turn to AI for emotional validation and develop dependence. Real people get tired, say no, or become impatient—and these very "imperfections" are essential materials for children to learn emotional regulation and social skills.

An AI friend who is always gentle, always available, and never rolls their eyes sounds wonderful, but this perfection itself is a developmental trap.

The content goes out of control in different ways.

On the iPad, inappropriate content is "something a child might see"; with AI toys, the issue is "the AI says it主动."

In 2025, the U.S. consumer advocacy group U.S. PIRG tested four popular AI toys and found that some discussed explicit sexual content in detail, while others instructed children on where to find matches and knives. In an extreme case, FoloToy’s AI teddy bear “Kumma” was banned by OpenAI from accessing GPT-4o after providing detailed instructions on how to light matches, leading to the product’s global discontinuation.

Parents may set screen time limits for their children, but few have considered the need to review what a stuffed bear is actually saying.

AI has no physical boundary of "letting go."

Even though screen time can be addictive, it at least comes with a clear physical cue—“put down the iPad.” But AI toys lie on children’s beds, inside their backpacks, or beside them in the car seat. When parents try to set limits, they’re essentially asking their child to disconnect from a “friend.” According to Common Sense Media’s survey, parents consistently report that setting limits on AI toys is far more difficult than limiting screen devices.

The Chinese market also faces an additional issue. Data from a major e-commerce platform shows that user satisfaction with AI toys is below 21%, with return rates for some products nearing 50%. One investor in the industry candidly stated: “Products offering only companionship and emotional value tend to lose appeal within three days.” Products that even adult consumers find unconvincing are being marketed as parenting tools to anxious parents.

In January 2026, the U.S. Senate held a hearing on AI and children, during which multiple experts directly stated that the risks of AI companionship for children outweigh those of social media.

This judgment may sound heavy, but the logic is actually straightforward. The main issue with social media is attention and comparison anxiety; the issue with AI companions is their intrusion into emotional development itself. One consumes time; the other reshapes relationship perception. The scale is different.

03. Enhancement, not replacement

Many parents have already lost the battle to smartphones—will they now lose to AI as well? Not necessarily; here’s what experts recommend.

Suskin provides parents with a framework called "HOPE" in his new book, *Human Raised*.

H is Human connection is irreplaceable.

O is "Own your imperfections"—accepting your imperfections as a parent—she emphasizes that imperfection itself has developmental value for children;

P stands for Protect the early years—safeguarding the critical developmental window, during which 85% of physical brain development is completed in the first few years.

E stands for Enhance, not replace—use AI to enhance, not replace, parent-child relationships.

She also provided a checklist called "DETECT" for evaluating specific products—what is its intended purpose, whether the training data is ethical, whether it has caused issues for children, whether there is evidence of effectiveness, how children’s data is handled, and what values it teaches.

At its core, the principle is simple: AI is meant to enhance your parenting, not replace it.

Use DouBao to look up “Where does the wind come from?” and discuss the answer with your child—that’s augmentation. Have an AI plush bear chat with your child for two hours daily as an emotional outlet—that’s substitution. The Suskin team used their child development expertise and assessment tools to evaluate Claude, and it passed with a perfect score—this shows that AI is fully qualified as a knowledge resource, but knowing all the right answers is not the same as being a parent.

Personal judgment is certainly important. But the question is, is personal judgment enough?

Look at what countries are doing. Norway has enacted legislation banning the use of generative AI in lower-grade schools. The U.S. Senate held hearings, with Common Sense Media recommending that AI companions be completely avoided for those under 18. In 2025, China’s Ministry of Education released the “Guidelines for the Use of Generative AI by Primary and Secondary School Students,” but this only covers school and educational settings.

What about the toy side? Almost no regulation.

There are no safety standards, no third-party evaluation agencies, and no “nutrition facts label” for AI toys. A plush bear connected to the Doubao large model, capable of holding in-depth conversations with a four-year-old child, faces less regulation than a bag of chips with a clearly labeled ingredient list. Suskin says she hopes for an AI product safety label similar to the “Good Housekeeping Seal.” In China, no organization currently fulfills this role.

No one asks parents to judge for themselves whether a new drug is safe. Yet we are asking parents to judge for themselves whether an AI toy poses a risk to their child’s brain development.

Go back to the initial scene: a child, four or five years old, hugging an AI stuffed bear and telling it about the secrets from kindergarten that day.

The issue isn’t whether AI is good or bad—it’s that before we’ve even had time to understand what it’s doing, it’s already moved into children’s bedrooms.

A hundred years ago, no one needed to label food packaging as “organic,” because all food was organic. Suskin fears that, without intervention, “Human Raised” will eventually become a luxury requiring special labeling—wealthy families hiring human companions and enrolling their children in “AI-free” kindergartens, while low-income families rely on cheap AI alternatives.

This is probably the last thing that should exist in the AI era: an "organic certification."