AI or not AI? A quick way to tell
Spam filters, face unlock, a 7:00 alarm, a bill-splitting calculator. One question sorts them, and it is the first idea of our Basic course.

"AI" gets stuck on everything from toothbrushes to spreadsheets, so it's easy to lose track of what the word actually means. Here's a quick test. For each situation below, decide: is AI at work here, or not?
- Your email moves ads and scam messages to the Spam folder on its own.
- Your phone unlocks when it recognizes your face.
- Your alarm rings at 7:00 every weekday.
- A calculator app splits the restaurant bill between friends.
Decide before you scroll on.
The answers
1. Spam filter — AI
Nobody wrote a rule for every possible scam. Spam filters learned from millions of emails that people marked as spam and worked out the warning signs on their own — odd links, certain phrases, suspicious senders. That's why they keep up with new scams that didn't exist when the filter was built.
2. Face unlock — AI
Face unlock learned what your face looks like from examples. No programmer wrote a rule for every angle, haircut, pair of glasses or lighting condition — the system learned to recognize you from data.
3. 7:00 alarm — not AI
It follows a fixed rule you set: same time, same days. Nothing is learned from data. It's useful, reliable software — but it isn't AI.
4. Bill-splitting calculator — not AI
It runs exact maths a programmer wrote. Same input, same answer, every time. That's ordinary software.
The one question that sorts them
Did a person write the rules, or did the system learn them from examples?
If a programmer wrote every rule — "if it's 7:00 on a weekday, ring" — it's ordinary software. If the system learned patterns from lots of data — "these are the signs of spam" — it's AI. More precisely, it's machine learning, the kind of AI behind almost everything that's called AI today, including AI chats like ChatGPT, Claude and Gemini, which learned language from enormous amounts of text.
Why the difference matters
It isn't just trivia. Knowing which kind of system you're dealing with tells you how to treat its output:
- Rule-based software is predictable. If the calculator is wrong, there's a bug, and it's wrong the same way every time.
- Learned systems are flexible but probabilistic. They handle situations nobody planned for — and they can also be confidently wrong. A spam filter sometimes catches a real email; an AI chat sometimes invents a fact.
That second point is the reason the most important AI skill isn't prompting — it's checking. We cover that in How to fact-check an AI answer in five minutes.


