How to Build an Educational Game with AI
- Published
- Updated
I built an educational game with Artificial Intelligence and published it online for free. The experience showed that AI can shorten the distance between a teaching idea and a working resource, provided that there is a clear learning objective and human validation at every stage.
The project reached nearly 200,000 people on LinkedIn, received more than 2,000 reactions and 130 shares. Educators in Portugal and Brazil took the game into their classrooms, parents played it with their children, and people asked for adaptations for adults and special education.
The numbers confirmed the interest. More importantly, the response showed that there is room to use AI creatively in education without giving up rigour.
What comes first: the game or the learning?
The starting point should always be the learning outcome. Before choosing tools, answer:
- Who will play?
- What should they know or be able to do at the end?
- Which mistake or doubt do we want to make visible?
- How will feedback be given?
- How long should each round take?
- Which accessibility needs should be considered?
Without these answers, it is easy to make something visually interesting that teaches nothing. AI accelerates production, but it does not replace pedagogical intent.
The four-stage process
1. Research and validate the content
I used Perplexity AI to structure the initial research. A research tool can help discover sources and organise questions, but every fact used in the game should be checked against references suited to the subject and age group.
Start with a table containing the concept, correct answer, plausible alternatives, explanation and source. This foundation makes the game easier to review.
2. Specify the game in a prompt
A useful development prompt should define:
- audience and learning objective;
- rules, scoring and end conditions;
- number and type of questions;
- feedback for correct and incorrect answers;
- behaviour on desktop and mobile;
- accessibility requirements;
- data that must stay outside the application.
The more concrete the specification, the less time is spent fixing decisions that were never made.
3. Build and iterate
I used Claude AI to develop the experience. Rather than asking for the complete game and accepting the first result, I worked in blocks: structure, logic, content, interface and corrections.
Test every change. A feature that looks correct in the code may fail with touch controls, on a small screen or when the user gives an unexpected answer.
4. Publish and observe
The game was published on Netlify. Before sharing, check:
- that it opens without unnecessary authentication;
- that it works at different screen sizes;
- that text is readable and buttons are clear;
- that it does not collect personal data without a need;
- that there is an easy way to report an error;
- that content and sources can be updated.
Publishing does not end the process. The questions and difficulties of the people who play reveal what should improve.
How to evaluate an AI-generated educational game
A simple evaluation can use five criteria:
- Accuracy: is the content correct and current?
- Learning: does the experience help people understand, or merely guess?
- Usability: can someone work out what to do without lengthy instructions?
- Inclusion: do the text, contrast, navigation and difficulty suit the intended audience?
- Privacy: does the game collect only indispensable data?
Ask a teacher, a subject specialist and several people from the intended audience to test it. Each group will identify different problems.
What this experience teaches us about AI in education
The main opportunity is not to generate more worksheets or more content. It is to let educators turn knowledge into experiences that previously required a technical team, test them faster and adapt the material to context.
When an experience includes variable answers or generated content, the principles of UX for Artificial Intelligence products also apply.
To explore this further, SuperHumano AI provides AI training for educational teams, while SuperHumano Academy offers interactive courses that show how a short experience can end with a concrete result. You can also explore tools by category in IAteca.
Frequently asked questions
Do I need to know how to code to build an educational AI game?
You can build a prototype without advanced experience, but you need to understand the basic logic, test what was generated and know when to ask for help.
Can AI create the game's questions?
It can help generate and adapt questions, but a specialist should verify accuracy, difficulty, sources and suitability for the audience.
Which tool should I choose?
Choose after defining the goal. Research tools, coding assistants and publishing platforms solve different parts of the process.
Context note: this article was published before the live session on 2 March 2026. The session has ended; the content was revised to preserve the method and make it reusable.