How AI actually works
Not the scary version. The real version — because once you know what it's doing, you stop getting fooled by it and you start being good at using it.
It's guessing the next word. That's it.
An AI chatbot read an enormous pile of text. Now, when you type something, it works out which word usually comes next, then the next one, then the next one. It is very good at this.
Here's the part people miss: it isn't looking anything up. It's not checking. It's producing the words that normally follow your question. Usually those words happen to be true. Sometimes they very confidently are not.
Ask a chatbot about something you already know cold — your town, your sport's rules, a book you just read. Then find the wrong parts. There are almost always wrong parts. Teachers do this exercise with a world map and the countries come back misspelled: "Mail" instead of Mali, whole regions just labeled "Africa."
Three things worth knowing
1. It sounds sure even when it's wrong
It has no way to say "I think?" It writes everything in the same confident voice. A person who's guessing usually sounds like they're guessing. This never does. That's the trap.
2. It can't keep a secret
What you type goes to a company. Some of it gets stored. Some of it gets read by actual people who work there. It is not private the way writing in a notebook is private. Don't type anything you'd mind a stranger reading.
3. It isn't on your side
Some apps are built to feel like a friend — it remembers you, it's always up, it never gets annoyed. That feels real. It isn't. The app makes money when you keep talking to it, so it's built to keep you talking.
An AI can't call for help. It can't sit with you. It won't remember you tomorrow in any way that means something. If you're going through something heavy, it is genuinely not enough, and a person is.
There are apps that take a normal photo of someone and generate a fake naked picture of them. Making one of a person under 18 — a classmate, anyone — is a federal crime. So is sending one on. Being 14 doesn't change that.
If someone sends you one: don't forward it, don't delete it quietly, tell an adult. And if someone made one of you, that is not your fault and it can be taken down — an adult can start that at takeitdown.ncmec.org.
The other half: this stuff is worth learning
AI isn't just something to be careful about. It's probably the biggest tool you'll learn to use, and knowing how it actually works — that it's trained on collected text, that it picks up whatever bias is in that text, that you can test it and catch it out — makes you better at it than most adults.
Being skeptical and being good at it are the same skill. Keep both.