I interviewed 12 families of only-child parents born in the 1990s. One thing kept coming up: they weren't teaching their kids — they were making up for what they themselves didn't have growing up.
Yuanyuan's mom (born 1990, Shanghai) bought her first AI learning tablet when her daughter was three and a half. She's an only child too.
"My mom bought me the complete set of Children's Pictorial and subscribed for three years. But no one taught me how to read it. In the end it all went to waste."
She bought the most expensive brand on the market — nearly 4,000 yuan. "I thought, nobody taught me when I was little. Now there's a machine that can teach my kid. I have to get the best one." Two months later, her daughter was using it to draw and watch cartoons.
This isn't an isolated story.
In Q1 2026, China's learning tablet market sold 1.252 million units. Despite a 1.0% year-on-year decline, spending by middle-class families on AI education hardware kept climbing. Homework Hero (Zuoyebang) captured 32.9% market share (up 4.6pp), with sales in the 2,000-3,000 yuan price range exceeding the combined sales of positions 2 through 4 (source: RotoTech, 2026-05-13). Xueersi followed at 20.1%, iFLYTEK at 18.5%.
But the real question isn't "should I buy one?" It's "what happens after I do?"
The Data Behind the Anxiety
76.3% of parents rank AI teaching depth as their top criterion when choosing a learning tablet (source: RotoTech, 2026-05-13). That number says one thing: parents know exactly what they're paying for — not hardware, but the machine's ability to teach.
NetEase Youdao's Q1 2026 AI subscription revenue surpassed 100 million yuan, up 70% year-on-year (source: Youdao Q1 2026 earnings report, 2026-05-21). Subscription models are becoming mainstream — parents are no longer buying a "device," they're buying a "service."
Meanwhile, problems are surfacing. In May 2026, Anhui Province Consumer Protection Commission released an evaluation showing that AI learning tablets make grading errors and produce formulaic essay comments (source: Anhui Consumer Protection Commission, 2026-05-27).
And a Ministry of Education survey (2026-06-01) revealed that 63.8% of children first encounter the internet before age 10, with 6.1% first going online before age 3.
The picture is clear: parents want the best — buy the most expensive device, subscribe to the most comprehensive courses. The result: devices gather dust, AI grading is unreliable, and children get hooked on screens too early.
The problem isn't "to buy or not to buy." It's "how to use."
AI Is a Ladder, Not an Elevator
Many parents treat AI education products like an elevator — put the child in, press a floor, and done. This is the biggest misconception.
AI is a ladder. It helps you reach things you couldn't otherwise reach. But you still have to climb.
What you need isn't "a machine that can teach my kid for me." It's a tool that makes the learning process visible, trackable, and adjustable.
Quick check: Ask one question to judge any AI product: Does it teach the child how to think, or does it just tell the child what the answer is? The former guides process. The latter reinforces outcomes.
The Three Screening Questions
If you're torn about whether to buy an AI education product, ask yourself three questions:
1. Does it teach knowledge or method?
- Knowledge ✗: "3+5=8" (child just memorizes)
- Method ✓: "How could you break down 3+5? Make a ten first, or split the smaller number?" (child needs to think through the process)
2. Does it give feedback or just a score?
- Score-only ✗: "3 wrong, 70% correct"
- Feedback ✓: "Question 2 had the right idea, but you skipped a calculation step"
3. Can you see the process?
- Black box ✗: Child answers → machine gives score → next question
- Transparent ✓: Logs every step → shows the thinking path → you can replay
Action tip: Before buying, spend 5 minutes running the three screening questions. If all three answers are ✗, it's probably a problem-drilling machine, not a learning tool.
The Three-Layer Companion Method
Here's a practical framework for putting AI products in the right place in family education:
Layer 1: Parent-Led (Core) | 15 minutes/day
In this layer, the human matters more than the machine.
- No AI: parent reads a story or works through a problem with the child
- Goal: build the understanding that "learning = someone is here with me"
- This is what AI cannot replace
Layer 2: Human-AI Collaboration | 20 minutes/day
In this layer, human + AI works better than either alone.
- Child practices with AI while parent observes the process
- Key observation: where does the child get stuck? How long does attention last?
- Turn AI's report into conversation: "How did you think through this problem?"
Layer 3: AI-Assisted (Free) | 3x/week, 10-15 minutes each
In this layer, the machine accompanies while the parent manages the boundaries.
- Child independently uses AI for review or preview
- Parent doesn't participate but knows what's being learned (weekly report, time distribution)
- Core mechanism: child does a "one-sentence summary" after each session — parent records it
Action tip: Stick a three-column chart on the fridge. Each evening, spend 3 minutes marking which layer you used today. After one week, you'll see your family's real AI usage pattern.
Replace Score-Tracking with Question-Tracking
The most common mistake parents make: treating AI products as a score tracker. "How many points today?" "Did you improve this week?"
This mindset makes "score" the endpoint of learning. A better alternative: pay attention to your child's questioning ability.
Weekly questioning checklist:
| Dimension | Beginner (1pt) | Intermediate (2pt) | Advanced (3pt) |
|---|---|---|---|
| Depth of questioning | Asks one "why" | Follows up 2-3 rounds | Turns it into a mini-investigation |
| Question variety | Only asks "what is" | Also asks "why/how" | Also asks "what if not" |
| Expression completeness | Keyword only | Complete sentence | Describes full context |
| Resource use | Only reads AI's answer | Asks AI for "more" | Cross-verifies with other tools |
Action tip: Friday evening, pull out a sheet of paper and look at the AI tool report together. Don't check scores — check whether the questioning level went up. Reward questioning, not scoring.
Learn AI Yourself First
This may be the most counterintuitive advice, but it's the most effective: learn to use AI yourself before teaching your child to use it.
Children imitate how you interact with AI. If you only type "what's the answer," your child will too. If you follow up, question, combine information — so will your child.
A three-week starter plan for parents:
- Week 1: Use AI for one work task — write an email draft, summarize meeting notes, look up a term. Then verify whether it's correct.
- Week 2: Ask AI an open-ended question (like "how would you explain photosynthesis to a 5-year-old"). Then follow up three rounds of "anything else?"
- Week 3: Let your child watch your complete AI workflow — ask → wait → verify → revise → finalize. Tell them: "AI makes mistakes too, so we check."
Action tip: Spend 5 minutes a day using a free AI tool for one small daily task. After one week, you'll have enough confidence to say "I know where it's good and where it fails."
Create a Weekly AI Review Ritual
Friday evening. 15 minutes. No scoring. Just conversation.
The 15-minute review template:
- First 5 min: "What did you use AI for this week?" (let the child talk, don't interrupt)
- Next 5 min: "What did you do well? What felt strange?" (guide expression)
- Last 5 min: "What do you want to try with AI next week?" (set a small goal together)
If your child says "AI skipped a step on this problem" — that's the best feedback. They're observing AI critically, which goes beyond "just finish it."
A Note to Only-Child Parents
You're anxious because you only get one shot. In multi-child families, there's a psychological "risk diversification" — if one child doesn't work out, there's another. Only-child families don't have that. Every expectation, every anxiety, every resource piles onto one person.
And that's exactly why you need to stay calm. Your child doesn't need an anxious companion. They need a calm observer.
You don't need to try every AI education product. You just need to find the one that fits your family's rhythm. You don't need to pile on "coding thinking" + "English immersion" + "logic training" before your child turns three. You just need to read them a book for 10 minutes every day.
Action tip for when anxiety takes over:
- Unfollow the product reviewers (they profit from your anxiety)
- Close your social feeds (you're not comparing education — you're comparing consumption)
- Open a blank document and write: "What do I want my child to be like in ten years?" The answer is probably not "accepted to Tsinghua" — it's "curious, and knows how to find answers." If that's the answer, most AI product evaluations need a reset.
What You Can Do Starting Tonight
| When | Action | Time |
|---|---|---|
| Tonight | Ask your child: "What's the one thing you really want to tell me today?" (instead of "did you finish homework?") | 10 min |
| This week | Run the "Three Screening Questions" on every AI device in your home | 20 min |
| This Friday | Do your first "weekly AI review" | 15 min |
| Next week | Learn one thing with AI yourself, then teach it to your child | 1 week |
References
- RotoTech. 2026 Q1 China Learning Tablet Market Report. 2026-05-13.
- NetEase Youdao. Q1 2026 Earnings Report. 2026-05-21.
- Anhui Province Consumer Protection Commission. AI Learning Machine Evaluation Report. 2026-05-27.
- Ministry of Education. 2026 National Survey on Children's Internet Usage. 2026-06-01.