AI always sounds sure – even when it is wrong
Why you should never blindly believe AI answers: invented facts, invented sources, hidden prejudices.
The tricky thing about AI: it always answers confidently – even when it is way off. Because it phrases things so fluently, even wrong answers sound convincing.
What you need to keep in mind
- It invents facts. Wrong numbers, dates and names, phrased convincingly. Experts call this "hallucinating".
- It invents sources. Sometimes the AI names books, quotes or links that do not exist at all. Always check for yourself.
- It is not neutral. It picks up prejudices and gaps from the texts it was trained on.
- "AI detectors" are not proof. Programs that supposedly recognise whether a text was written by an AI are often wrong.
Check it. With every AI answer, ask yourself: is that really true? Open real sources, compare with what you already know. Only if you know something yourself will you notice when the AI is talking nonsense.
A good test: ask the AI about your school or a person from your town. You will often get confident answers that are completely made up.
An AI gives you a year and a source for it. What do you do?
AI can invent facts and even sources and sounds convincing while doing it. Check numbers and actually open the source – confidence is not proof.