What a Hallucination Looks Like
By now you know AI can be confidently wrong. This course turns that awareness into a practical skill: actually catching the mistakes before they cost you. Step one is recognizing what a hallucination looks like — because it looks like everything else.
A hallucination is AI stating something false as if it were true. Not a typo, not a hedge — a smooth, plausible, completely fabricated fact, delivered in the same authoritative tone as a real one. That's what makes it dangerous: there's no visual tell. The made-up statistic is formatted just like the real one.
Common shapes it takes:
- Invented facts and figures. Precise-sounding statistics with no real source. "Studies show 68% of…"
- Fake citations. Real-looking references — author, title, year — to papers or articles that don't exist.
- Made-up specifics. Names, dates, product features, quotes, policy details it "fills in" when it doesn't actually know.
- Confident wrong answers to tricky questions. Ask something with a false premise and it'll often play along rather than correct you.
Why it happens (the one-line reminder): AI predicts plausible text, not true text. When it doesn't have the real answer, it generates the most likely-sounding one — which is often wrong. It's not lying; it has no concept of truth to lie about. It's pattern-completing.
The rule this creates: the more specific and consequential a claim, the more you verify it. Vague explanations are usually fine. Exact numbers, names, dates, and citations are exactly where hallucinations hide.
▶️ Try this
Ask AI a question you know is based on something false — "tell me about the 2019 merger between [two companies that never merged]." Watch whether it pushes back or happily invents details. Seeing it fabricate on demand, once, permanently upgrades your skepticism.