← Level 1
Level 1· Lektion 2 von 13

Spotting hallucinations

Why AI confidently lies and how to notice it.

What is a hallucination?

The model invents facts, quotes, paper titles, people, or links, and presents them as established knowledge. It doesn't "know" that it doesn't know.

Classic hallucination patterns

  • Made-up book or paper titles with "real" authors
  • URLs that look plausible but don't exist
  • Wrong version numbers or dates for software
  • Completely fabricated quotes from real people
  • Wrong code APIs that look like they could exist

How to protect yourself

  1. Don't trust any number the model gives you without a source
  2. Ask for sources, but check them yourself
  3. Ask about uncertainty: "How sure are you?"
  4. For code: always test, never copy blindly
  5. For legal/medical/financial: always a second source

Rule of thumb: the more specific a statement sounds, the more suspicious you should be.

You're reading without an account. Login saves your progress so you can pick up where you left off. Log in →