Two questions come up in almost every conversation I have with parents now. The first: if AI can already write code, is there any point in my child learning to? The second, usually straight after: my child is nine, or fourteen — what should they be learning now?
They're good questions, and the answer to the second one changes a lot with age. Here's how I'd think about AI and coding for kids and teens, stage by stage.
Quick Answer: Coding still matters, because AI writes code that often looks right and isn't — and only someone who can read code can tell the difference. For kids aged 7–9, start with coding alone and leave AI out of it. From 10–12, add real typed code and an honest explanation of how AI works. From 13–17, teens should be building something that behaves like AI, not just prompting one. Be sceptical of any class where "learning AI" means writing prompts.
Does Coding Still Matter If AI Writes the Code?
Yes, and the reason is narrower than the usual "it teaches critical thinking" answer.
AI coding tools are genuinely good. They are also confidently wrong on a regular basis. They produce code that runs, looks reasonable, and quietly does the wrong thing. The person who catches that is the person who can read the code and hold the whole problem in their head. The person who can't is stuck — they can't tell a working answer from a broken one, so every output is a coin flip.
That's the real divide forming, and it isn't between people who use AI and people who don't. It's between people who can check AI's work and people who can only accept it. Learning to code is how a child ends up on the right side of that line.
There's a second reason that matters more for younger kids: debugging. Sitting with something broken, forming a guess, testing it, being wrong, and going again is a genuinely hard habit to build. It doesn't come from watching a model produce an answer in two seconds.
Ages 7–9: Coding First, Leave AI Alone
At this age I'd skip AI almost entirely, and I say that as someone who teaches it.
A seven-year-old is still building the foundation that makes AI comprehensible later: the idea that a computer does exactly what you tell it, in order, and that when something goes wrong it's because of an instruction you wrote. That mental model is what everything else sits on. Introducing a system that appears to think, at the exact moment a child is learning that computers don't, muddies the one lesson they need most.
What works well instead:
- Block-based coding such as Scratch, where the logic is visible and nothing breaks because of a missing semicolon.
- Building games they actually play. This is why Roblox coding for kids works so well at 7–12 — the feedback is immediate and the motivation is already there.
- Finishing things. A small game that works, shown to someone, beats a big one abandoned halfway.
If your child asks what AI is, answer honestly — our guide to AI for kids has a version that works at this age. Just don't build the coding curriculum around it yet.
Ages 10–12: Real Code, and an Honest Look at AI
Somewhere around 10, most kids are ready for two shifts at once.
The first is moving from blocks to typed code. It's a real step up — spelling counts, punctuation counts, and the computer stops being forgiving. Handled well, this is where a child learns that an error message is information rather than failure. Python is the usual landing spot, and scripting in Roblox with Luau is a gentle route into it because the payoff is still a game.
The second shift is that they can now handle a real explanation of AI instead of a metaphor. Not "it's like a robot brain" — closer to: it has read an enormous amount of text, and it is very good at guessing which word comes next. That one sentence explains most of what they'll observe, including why it makes things up. A ten-year-old can hold that idea comfortably, and it inoculates them against both unhelpful extremes: that AI is magic, or that it's always right.
This is also the age to be clear that using AI to do homework isn't the same as learning. Kids work that out faster than adults expect, once someone explains what the tool is actually doing.
Ages 13–17: Build AI, Don't Just Use It
Teens are ready for the thing most "AI for teens" classes skip entirely: building one.
There's a wide gap between a teenager who can write good prompts and one who understands why a language model produces what it does. The first is a skill with a short shelf life — interfaces change, and prompting gets easier every year. The second is durable, and it's the part that carries into university and work.
In practice, that means teens should meet ideas like these in code:
- Next-word prediction: building something that learns what tends to follow what, then watching it generate text that's fluent and completely wrong.
- Probability and randomness: why the same prompt gives different answers, and what "creativity" actually means in a model.
- Tokens and context: what the model can really see, and why it forgets.
- Meaning as numbers: how words become vectors, and how a machine ends up with a usable sense of similarity.
- Knowing when not to trust it: where hallucination comes from structurally, rather than as a warning label.
That's the shape of our AI & Python course for ages 13–17: Python first, then building a working text generator from scratch over 12 lessons. Students come out able to explain why ChatGPT sounds so certain while being wrong — because they've built the small version and watched it happen.
Teens who'd rather build something visible have other routes into the same rigour. Web development puts their work on the live internet and into a portfolio, and game development with Construct 3 teaches event-driven logic through games. Both pair well with AI afterwards.
How to Tell a Real AI Class From a Prompt-Writing Workshop
"AI classes for kids" now covers an enormous range of quality, and the labels don't help. Three questions sort it out quickly:
- Does the child write code, or only prompts? If the whole class happens inside a chat box, it's a tools workshop. Useful for an afternoon — but it isn't teaching AI.
- Can they explain why AI gets things wrong? A child who has built a tiny model can tell you. A child who has only used one can't.
- What do they take home? Something they built and can demo beats a certificate.
The same instincts apply to choosing any provider — small classes, real projects, and an instructor who can go off-script when a student asks something unexpected. We've written about that in more depth in our guide to choosing coding classes for kids in Singapore.
Where to Start
If your child is 7–12 and new to this, start with coding and let AI wait: Roblox Game Dev for 7–12, then Python from around 12. If you have a teen, AI & Python is the direct route — Python foundations first, then building a language model small enough to understand completely.
Classes run in person at our Jurong East studio, a short walk from the MRT, or live online from anywhere in Singapore, in groups of up to 12. Every course starts with a free 45-minute trial lesson, which is genuinely the fastest way to find out whether it suits your child. Book a free trial class and see what they make.
