AI School › How it works · Lesson 8 of 130

🎭 Why it sometimes fails: hallucinations

AI can invent facts with great confidence. These are called hallucinations.

Sometimes AI answers with invented information, delivered in a very confident tone: wrong dates, books never written, people who never existed. Experts call these "hallucinations". It happens because the AI completes sentences in the most likely way: if the true answer is rare and a false one "sounds better", it may pick the false one.

It is like a classmate who wants to answer at all costs, even when they should have said "I have no idea". The golden rule: important information always gets verified from another source, a book, a reliable website, a knowledgeable adult. AI is a great starting point and a terrible finish line.

🎒 Try it in class (or at home)

Detective test: ask the AI for the biography of a made-up person ("Who was Gianfranco Meravigli, the explorer?"). If it answers with details, you have just seen a live hallucination.

⚠️ Watch out

If an AI answer concerns health, money or safety, always check it with an adult before using it.

← Lesson 7: The trick: predicting the next word · Lesson 9: The prompt: your words matter →

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.