The simplest way to teach kids how AI works (no screen needed)

I wanted a way to explain AI to my students that didn’t require opening a laptop. Something they could feel, not just hear. This activity takes eight minutes, needs zero technology, and reliably produces that “oh, that’show it works” moment.

The game: One Word Story

Everyone sits in a circle and builds a story together — but each person can only say one word at a time. You go around until the story reaches a natural end (or collapses into laughter, which happens more often).

A typical round sounds like:

“Once… there… was… a… very… hungry… robot… who… wanted… pizza…”

The stories get weird fast. Someone says “invisible” when you expected “small,” and the whole direction shifts. That surprise is the point.

The moment that changes everything

After the game, I ask: “What were you actually doing when it was your turn?”

After a beat, someone lands on it: “I was thinking about what word made sense next.”

Exactly. Each person was reading everything that came before, and choosing the word that felt most natural to follow. Predicting, one step at a time.

Then I tell them: That’s what ChatGPT does.

Not metaphorically. Literally. A language model reads all the text before a gap and predicts the most likely next word — one at a time, in sequence. The technology behind it is more complex, but the core behaviour is exactly what they just acted out.

Why it works

The kids don’t just hear about AI — they become the process. When the concept lands, it lands in their bodies, not just their heads.

It also makes the limitations obvious without any lecturing. I ask: “Did your story always make sense?”They laugh and say no. That’s the nature of prediction — “most likely” isn’t the same as “correct” or “true.” This is why AI sometimes sounds confident but gets things wrong.

Eight-year-olds understand this immediately after One Word Story. They lived it.

Try it at home

You just need a dinner table and at least two people. Play a round, then ask: “What were you doing when you picked your word?”Let them arrive at “predicting” themselves. Then tell them that’s the engine inside ChatGPT.

A few good follow-up questions:

• If AI is always predicting the next word, can it ever know something is true?
• Why do you think it sometimes gives different answers to the same question?

These build the kind of AI literacy that actually lasts.