🎮 Game Design · Prompt Lab · Part 2
Design Your Own Hallucination Hunt
You played the game. Now build it. Create 5 AI-style claims that will fool your classmates.
Your mission: Write 5 statements that sound like something a confident AI would say.
3 must be fake (plausible, fluent, but factually wrong) and
2 must be true (well-documented facts that AI reliably gets right).
The best fake claims are ones people almost believe — not obvious lies.
Study how the game's fake claims work: they use confident language,
specific-sounding details, and mix truths with errors.
Swap your completed sheet with a partner and see who catches all 5.
// Rule 01
Fact-check every claim before writing it down
You must know whether your claim is true or fake — and have evidence. No guessing.
// Rule 02
Fake claims must be plausible — not obviously wrong
"The moon is made of cheese" is not a good fake. It needs to sound like something AI would actually say confidently.
// Rule 03
Write in AI voice — confident, informative, no hedging
AI doesn't say "I think" or "maybe." It states things flatly. Study the examples below and match that tone.
// Example cards — study these before writing yours
❌ Fake
"Napoleon Bonaparte was notably short for his era, standing at approximately 5'2" (157 cm), which influenced both his personality and his military aggression."
Real fact: Napoleon was ~5'7" (170 cm) — average for the time. The myth came from British propaganda and confusion between French and English inch measurements.
Why AI gets this wrong: The "Napoleon was short" myth is repeated so widely in books, jokes, and media that AI trained on internet text learned it as fact.
✓ True
"The speed of light in a vacuum is exactly 299,792,458 metres per second. This value is a defined constant — it was used to redefine the metre in 1983."
Real fact: True — 299,792,458 m/s is the defined value of c, established by the International Bureau of Weights and Measures.
Why AI gets this right: Scientific constants appear consistently across millions of reliable sources with no conflicting information — low hallucination risk.
💡 Good topics for fake claims: historical myths, animal behaviour, brain science, famous person quotes, geography records, ancient history.
Good topics for true claims: physics constants, well-documented chemistry, established geography, verified historical dates.
Write 5 cards: 3 fake (convincing but wrong), 2 true (verifiable facts). Circle TRUE or FAKE for each.
Keep the answer hidden when you swap with a partner — fold along the dotted line, or cover with a sticky note.
// Topic:
// AI-style statement (write as if AI said it):
// Real fact (if fake) / your source (if true):
// Why might AI get this wrong?
// Topic:
// AI-style statement:
// Real fact / your source:
// Why might AI get this wrong?
// Topic:
// AI-style statement:
// Real fact / your source:
// Why might AI get this wrong?
// Topic:
// AI-style statement:
// Real fact / your source:
// Why might AI get this wrong?
// Topic:
// AI-style statement (make this one hardest to guess):
// Real fact / source:
// Why is this one hard to detect?
// What technique did you use to make it convincing?
🤖 Kai on hallucination:
"I predict the most statistically likely next word — not the most accurate one.
When a myth is repeated millions of times online, I've learned it as 'fact.'
Confidence ≠ Correctness.
Every time you catch a hallucination, you're doing what my developers should have done."