A GUIDE FOR FAMILIES AND TEACHERS
How to make games with AI without turning the child into a spectator
A child does not need to begin by memorising syntax to build a first game. They can describe an idea, test what the AI produces and learn to direct each improvement. The point is not for the machine to do more, but for the child to make more decisions.
By Fernando Angulo · Published 4 August 2026
Making games with AI means using natural language to design, generate and improve a digital game. When the process is done well, the idea, rules, tests and decisions remain the child's; the AI accelerates the writing of code.
What vibe coding means for children
Vibe coding means describing what a program should do to an AI, then working on the result through instructions, tests and corrections. For a child, it shortens the distance between “I want to make a game” and “I can play something I imagined”.
It does not mean accepting everything the AI returns. The child should check whether the game works, explain what is wrong, decide what changes and recognise what they do not yet understand. That conversation with the result is the learning.
How to make a game with AI, step by step
- Choose an original idea, such as an octopus collecting rubbish before time runs out.
- Define one initial rule: what the player controls, how they win and how they lose.
- Request a small first version that runs in a browser without accounts or external services.
- Play before adding anything. Write down one fault or one specific improvement.
- Change one thing and test again, so each instruction can be connected to its result.
- Explain and show the game: what the child decided, what the AI did and what needed correcting.
What a child learns while making
Making a game requires turning a vague intention into observable rules. The child practises problem decomposition, sequences, conditions, hypothesis testing and precise communication without meeting those ideas first as a theoretical list.
They also learn an important limit: a convincing answer can still be wrong. If the character walks through a wall or the score adds up incorrectly, the AI is not the authority; the test is.
What a safe environment requires
- A present adult who can see the instructions and results.
- An account and key managed by the adult or school, never by the child.
- Filters before and after generation to prevent unsuitable content and unsafe code.
- An isolated preview without forms, trackers or connections to external domains.
- A way to save or export the game without publishing personal data.
What AI does not replace
AI does not replace the child's idea, their judgement about whether something works or the responsible presence of an adult. Nor does every session with a tool become learning: copying a prompt and accepting the first result is consumption through a different interface.
The goal is not to produce the most complex game. It is for the child to explain what they wanted to build, what went wrong, what they changed and why the final version is theirs.
Questions about making games with AI
- What age is suitable for making games with AI?
- From around age nine, many children can make simple games with AI when the format matches their reading, writing and ability to sustain an idea. Ages 9–10 benefit from short or spoken instructions; from 11–12, children can iterate rules and visual design more independently. Age is a guide, but observable autonomy matters more than a birthday.
- Do children need to know programming before they start?
- They do not need programming knowledge to create a first version, but they do need to learn how to describe, test and correct. Code stops being the entrance barrier and becomes something they can begin to read when they need to understand a fault. Programming does not disappear; it arrives in a different order.
- Does making games with AI teach real programming?
- It can teach computational thinking and programming concepts when the child makes decisions and tests each change. It does not do so when they merely request a complete game and accept the result. A useful rule is to make one change, run one test and explain what happened.
- Is it safe for a child to use an AI tool?
- Only in an environment designed for children and with adult supervision. The child should not open their own account or share personal data, and generated code should run in isolation without unrestricted Internet access. Safety depends on configuration and facilitation, not a product label.
- What kind of game should a child make first?
- Start with a one-screen game, one objective and a simple winning rule. A maze, collecting game or obstacle-dodging game can be tested quickly and improved one thing at a time. A small completed project teaches more than an enormous abandoned idea.
- How is vibe coding different from using Scratch?
- Scratch builds programs from visible blocks; vibe coding directs creation through natural language and evaluates code generated by an AI. Both reduce barriers, but expose different parts of the process. They can complement each other: Scratch explains structure while vibe coding accelerates the journey from an original idea to a prototype.
Their first creation should be their own
At Vibe Coders Club, children build games and tools from their own ideas, with an adult beside them and no AI accounts in their name.