Every week, a new enrichment centre in Singapore adds "AI" to its programme name. Some of them mean it. Many of them don't.
This post is for parents who are genuinely trying to figure out what AI education for children actually looks like — what's worth your money, what's marketing, and what builds something real in a child's mind that lasts beyond the school holiday.
We'll be direct, even about our own industry.
What "AI for Kids" Often Actually Means
Let's start with what to be sceptical about.
The child uses a visual coding platform (e.g. Scratch or Makecode Arcade) to build a simple game or animation, and the word "AI" appears in the marketing because the platform has a drag-and-drop "machine learning" block somewhere. This is fine as an introduction to computational thinking. It is not AI education.
The class talks about AI — what it is, how it works, ethical considerations. This is valuable. But if the child leaves without having used AI to build something, the knowledge is abstract. It won't stick the way hands-on learning does.
The child types prompts into ChatGPT, gets some outputs, pastes them into a Word document. The lesson has the word "prompting" in it. This is not the same as learning to build with AI.
None of these are fraudulent. Some are genuinely useful. But they're not what we mean when we talk about AI education that changes how a child sees themselves.
What Actually Works
The research on children's learning is unambiguous on one point: learning sticks when it's attached to making something real.
This is why children who learn to cook remember more about chemistry than children who read about chemistry. It's why children who build robots remember more about physics than children who study physics in the abstract. The production of a real object — something that didn't exist before and now does — creates a memory anchor that worksheets and lectures simply don't.
For AI education to genuinely work with children aged 9–12, it needs to satisfy a few conditions:
- The child must be the agent, not the audience. They use the AI tool. They make the decisions. They observe what works and what doesn't. The instructor guides but does not build for them.
- The output must be real and shareable. A slide deck about an app idea is not the same as the app. A working, deployed application that the child can send a link to is a fundamentally different kind of achievement.
- The project must be chosen by the child. Motivation matters enormously in children this age. A child who is building something they actually care about — a fan site for their favourite manga series, a quiz about their obsession with Pokémon characters — will work harder, iterate more, and remember more.
- There must be a moment of public sharing. When a child has to explain what they built and how, they consolidate their understanding in a way that private completion never achieves. The audience doesn't need to be large. It just needs to be real.
What's Age-Appropriate for 10–12 Year Olds?
This is a common parental concern, and it's worth addressing seriously.
Children aged 9–12 are in what developmental psychologists call the concrete operational stage transitioning toward formal operational thinking. They're capable of logical reasoning about real objects and events, and starting to handle more abstract "what if" thinking. This makes them ideal candidates for AI-assisted building — they can think in systems, handle cause-and-effect loops, and take pride in real output.
They should not be using general-purpose large language models without supervision. Not because AI is inherently dangerous for this age group, but because open-ended chat tools can produce outputs that require adult judgment to evaluate. Good AI education for this age requires purpose-driven and monitored guidance and coaching that produce a specific type of output (a website, a game, an app) rather than open-ended text generation.
That is why at Dreambuilders.sg we emphasise in-person sessions with small group-to-coach ratios so that each child is guided and supported on their journey of building with AI.
The One Question to Ask Before Enrolling
Before spending money on any AI or coding programme for your child, ask the provider this:
"What will my child walk away having made?"
If the answer is a certificate, a workbook, or a vague gesture toward "skills," dig deeper. If the answer is a specific, deployed, shareable thing — a website, an app, a game — with a clear process for how they got there, that's a programme worth considering.
The measure of a good AI programme for kids isn't what your child learned about AI. It's what your child built using AI.