Bot or Not: the quiz
Each question has two answers: one a person, one a bot. Round 1 the bot writes like a chatbot. Round 2 it tries to blend in.
Honest label: we wrote every answer on this page to show the habits, so this is a warm-up. In the daily game the humans are real players and the bot is a real AI.
Round 1: easy
The bot writes like a chatbot. You'll get most of these.
1. What did you have for breakfast?
Real breakfasts are boring and badly described. "Wholesome", the trio of toppings and the full sentence are chatbot habits.
2. What is 17 × 23?
People asked to do maths in their head guess or give up. Instant exact maths, no working, is the giveaway.
3. What can you hear right now?
A real "now" is specific and a bit annoyed. "Gentle hum" and "symphony" are pleasant and could be anywhere.
4. Cats or dogs?
People pick a side. Seeing both sides, with an exclamation mark, is the eager-assistant voice.
5. What was your first phone?
Real memories are patchy and end abruptly. Invented ones come with a little lesson at the end.
6. Describe the colour blue to someone who has never seen it.
Asked something hard and odd, people shrug. A polished, quotable answer is suspicious.
7. What's the most overrated food?
Hedging ("subjective", "many people feel") instead of an opinion is a classic tell.
8. Spell "definitely" backwards.
Too fast and too sure. The human checks themselves out loud.
Round 2: hard
Now the bot has been told to sound casual: lowercase, "lol", short answers. The obvious habits are gone. Look for the subtler one: it picks the most common answer.
1. What did you have for breakfast?
Disguised bots still answer exactly what was asked, with a filler phrase ("nothing fancy"). People often dodge or bend the question.
2. What is 17 × 23? No calculator.
A bot told to be human fakes a guess with a stock reaction ("ugh math lol"). The human questions the premise, which bots rarely do.
3. What can you hear right now?
Cars and a fridge are the most likely sounds in the world. Real "now" details are odd and particular: "its weird clunk thing".
4. What was your first phone?
Nokia plus Snake is the most common answer there is. A specific model and a strange detail are more human.
5. Cats or dogs?
Even disguised, the bot can't help seeing both sides ("cats are cool too tho"). People's answers are lopsided, or off the menu entirely.
6. Describe the colour blue to someone who has never seen it.
The bot still produces a neat metaphor, just in lowercase. Faced with a hard, odd question, plenty of people simply skip it.
7. What's the most overrated food?
Avocado toast is the internet's stock answer. Real opinions come with a personal grudge and a specific detail.
8. What's a hill you'd die on?
An internet-famous debate is the safest pick, so it's the bot's. People's hills are local and specific.
What round 2 teaches
- Common beats specific. A bot reaches for the likeliest answer (Nokia, avocado toast, pineapple on pizza). People's answers are oddly particular.
- It answers the question asked. People dodge, bend the question, or skip it.
- Hedges creep back. "cats are cool too tho" is a disguised "both are great!".
The daily game's bot is told to blend in, like round 2, and you choose the questions. More on picking them in questions that catch an AI.
Now try it against a real AI →
Questions
Are these real human answers?
No. We wrote all of them to show the typical habits. The daily game uses real players' answers and a real AI.
Why is the quiz easier than the game?
Round 1 shows the habits clearly. Round 2 is closer to the game, where a real AI is told to blend in, and you choose the questions to make it slip.
What is the best tip for the hard round?
Ask yourself which answer is the most common one anyone could give. That is usually the bot.