Exercise 3: The Animal Guessing Game

You know the drill. Think of an animal — any animal — and keep it to yourself. The AI gets 20 yes-or-no questions to figure out what it is.

This is a classic car ride game, and AI turns out to be surprisingly good at it. It works through the possibilities the same way a sharp player would: broad categories first, then narrowing down. Mammal or not? Lives on land or water? Bigger than a dog? Could it beat you in a fight?

How to play

Step 1: Think of an animal. Don't tell it yet.

Step 2: Copy the prompt below and paste it into any AI tool. Hit send.

Step 3: Answer each question with Yes, No, or Sometimes. The AI will keep count and make its guess when it's confident — or when it runs out of questions.

Prompt

Let's play 20 questions. I'm thinking of an animal. You have 20 yes-or-no questions to guess what it is — ask them one at a time and keep count. [difficulty] Make your best guess when you're confident, or when you've used all 20 questions.

difficulty

The harder you make it, the more fun it gets. Common animals — dog, cat, elephant — it'll usually nail in 8–10 questions. Pick a quokka, an axolotl, or a tardigrade and watch it work harder. If it guesses wrong, tell it what you were thinking and ask how it would approach that one differently.

What this shows you

The guessing game is a surprisingly good window into how AI reasons. A few things to notice as you play:

It deduces, it doesn't guess randomly. Watch how it uses your answers to eliminate possibilities — if you say it doesn't live in the ocean, it immediately rules out an entire category. That's logical reasoning through elimination, the same thing a good 20-questions player does.

It tracks context across the whole conversation. By question 15, the AI remembers everything you've said in questions 1 through 14 and is using all of it. This is the same skill that makes it useful for long, complex tasks.

It knows when it's confident vs. uncertain. A good AI player will tell you when it thinks it has the answer vs. when it's still narrowing down. That calibration — knowing what you know and what you don't — is one of the harder things to get right, for humans or AI.

Keep playing

Once it guesses (right or wrong), try these:

Switch roles — tell the AI to think of an animal and you ask the questions. Ask it to answer honestly and reveal what it was thinking at the end.
Try other categories — the same game works with famous people, movies, or foods. Just change the first line of the prompt.
Play with kids — this is a great one to hand off. Let them think of the animal and read the questions out loud.