Ages 8–10 and AI: Guided Learning and Creative Projects

Somewhere between eight and ten, a child stops needing you to read the screen and starts needing you to argue with it. They can type a question, follow an answer, and build something that takes more than one afternoon. What they cannot yet do is tell a confident wrong answer from a right one.

That gap defines the whole stage. The tools below all assume a fluent reader with an adult within earshot.

Six worth your time

Khanmigo

Khanmigo, the tutor from Khan Academy, refuses to hand over answers. Ask it to solve a fraction problem and it asks your child what they would try first. Parent and teacher dashboards show the full conversation history, which removes the guessing about what happened while you were making dinner.

Best used for the subject your child finds hardest, at the point where frustration usually ends the session.

Machine Learning for Kids

Machine Learning for Kids connects model training to Scratch, which most kids this age already know from school. Your child trains a classifier on examples they choose, then drops it into a Scratch game they built. A project might take three sittings. The payoff is a working thing they made.

This is where “AI learns from examples” turns from a sentence into an afternoon. When their model keeps calling every photo a dog, they have to go back and fix their training data, which is the same problem professionals spend their careers on.

Google Teachable Machine

Shorter than the above, and no coding needed. Teachable Machine trains an image, sound, or pose model in the browser in about ten minutes. Show it two hand gestures, label them, watch it learn. Nothing leaves the machine by default.

Pair it with a question your child can test: does the model still work if I stand somewhere darker? Does it recognize my sister?

Canva for Education

Canva for Education puts AI image and writing features inside a real project, so the poster or comic stays the point and the AI stays a supporting tool. Teachers can switch individual AI features off, which makes the classroom version the one to want. Families get the free tier through a teacher.

Book Creator

Book Creator lets kids build a real book with pages, audio, and drawings, then publish it as a link they can send to a grandparent. Kids at this age want an audience. Giving them one does more for their writing than any prompt trick. See our story workshop for a session plan.

Seek and Merlin

Seek identifies plants and insects from a photo; Merlin identifies birds from their song. Both stay free of ads and accounts, both send kids outdoors, and both give you a natural moment to ask the best question in AI education: how sure do you think it is?

The two habits to build now

Check one thing. Every session, pick one claim the AI made and verify it somewhere else. A library book, a second search, a parent who knows. Kids who build this reflex at nine carry it into thirteen, when nobody is watching the screen. Our guide on teaching kids that AI can be wrong has scripts that work at this age.

Say what you used it for. Not as a confession, as a normal part of describing a project: “I asked it for three title ideas and picked one.” Kids who narrate their AI use keep control of it. Kids who hide it lose track of where their own thinking stopped.

Where the line sits on schoolwork

Ask your child’s teacher, then hold the line at home. A rule that works across most classrooms: AI can help you understand, plan, and check. It does not write the sentence you hand in. If your child cannot explain a paragraph they submitted, it was not theirs.

What still needs you in the room

General chatbots, image generators with an open prompt box, and anything with a chat-style persona. Eight-year-olds are old enough to type anything into a box and young enough to believe what comes back. Keep the accounts yours, keep the screen visible, and revisit what your child should never type into an AI once a term.

By eleven or twelve, the balance shifts again. Our guide for ages 11 to 12 covers research, coding, and the run-up to their own accounts.