Wovi

Free AI projects for kids · Ages 9–13

Build one thing.
Test it. Explain what changed.

A useful AI project is more than opening a chatbot. These two guided browser projects let children change the data or rules, rerun the same world and finish with an evidence-based explanation.

Choose a free project See the connected world map

Wovi project guide · Updated August 13, 2026 · No account required

PROJECT LOOP

Predict → build → test → explain

QUESTION→ONE CHANGE→SAME TEST→EVIDENCE

The project ends when the learner can explain why the second result changed—not when the screen says “complete.”

BEST FORAges 9–13Beginners and Scratch learners
TIME5–10 minutes eachOne complete project loop
SETUPWeb browser onlyNo download or child account
ADULT ROLEAsk for evidenceNo AI expertise needed

Choose by what the learner changes

Two beginner projects.
Two different kinds of AI.

Start with examples and classification, or start with an agent policy and Python. Both use a fixed world so the effect of one change stays visible.

PROJECT 01 · MACHINE LEARNING

Reef Rescue classifier

Train a transparent nearest-neighbor model to flag reef risk, then improve its evidence when unfamiliar cases expose blind spots.

  1. Predict: Which examples will the model need?
  2. Build: Choose a labeled training set.
  3. Test: Compare false alarms and missed risks.
  4. Explain: Why did the varied set improve?
5–10 minNo codingFree online
Start the ML project Read the machine-learning guide →
PROJECT 02 · AI CODING

Agent Maze policy

Predict a scout's route, diagnose why a reward rule favors a dangerous shortcut and reveal the corrected policy as Python.

  1. Predict: Which route will score highest?
  2. Build: Set a visible safety penalty.
  3. Test: Run the same deterministic maze.
  4. Explain: Why did the route change?
5–8 minBlocks + PythonFree online
Start the coding project See the block-to-Python bridge →

A simple project record

Capture thinking
without creating an account.

A parent, teacher or learner can use four prompts on paper or in a local document. Do not include a child's identifying information.

  1. 1
    My prediction

    I think the model or agent will… because…

  2. 2
    The one change

    I changed this example or rule…

  3. 3
    What the test showed

    The fixed test behaved differently by…

  4. 4
    What remains uncertain

    One case I would test next is…

PROJECT SUCCESS

Not “the AI got it right.”

Observable: the learner names the data or policy value that changed.

Explainable: the learner connects that change to a result in the same test.

Critical: the learner identifies a limit or next case rather than declaring the system finished.

This evidence is more useful than completion time alone.

Choose the next route

Project first.
Then deepen the concept.

Teach AI with a 20-minute lessonUse one of these projects with prediction, testing and reflection prompts →Free AI games for kidsCompare the two playable missions and their learning outcomes →Continue across connected worldsKeep completed missions and fixed capability badges only in this browser →

AI projects, clearly explained

Questions from families and educators

What are good AI projects for kids?

Good beginner AI projects let a child change data or rules, test the same case again and explain the result. Wovi currently offers a classification project in Reef Rescue and an agent-and-Python project in Agent Maze.

Are these AI projects free and online?

Yes. Both projects run free in a browser with no download or child account. A desktop or tablet is recommended, and each core project takes about five to ten minutes.

Does a child need coding experience?

No. Reef Rescue starts with labeled examples and Agent Maze starts with visible reward rules. Agent Maze then reveals the matching Python so coding becomes a way to describe an idea the learner has already tested.

Can these projects be used for school or homeschool?

They can support a short guided activity for ages 9–13. Ask the learner to record a prediction, one change, the result and an evidence-based explanation. Educators should review local curriculum and safeguarding requirements.

What does the learner make or produce?

The current prototypes do not export a file or public profile. The meaningful output is a completed model-debugging loop: a prediction, a changed training set or policy, test evidence and a clear explanation of what improved and what remains limited.

Founding families

Want the next free AI project?

Join the adult early-access list for new project worlds, curriculum notes and optional family research invitations.

Join early access Free to join · Parent email only · Unsubscribe anytime