Why AI makes things up

It predicts likely words. It doesn't look things up. Everything follows from that.

The lesson

Hallucination in plain English, with real examples: invented citations, fake case law, plausible-but-wrong history. Why the machine can't tell you when it's guessing — and why the dangerous errors aren't the absurd ones, they're the plausible ones.

Try it — 60 seconds

Ask AI for 3 published sources on [a niche topic you know]. Then actually look them up. Notice how real the fake ones sounded.

Copy it into ChatGPT, Claude, or Gemini — whichever you already use. Fill in the brackets with your real life.

Watch out

It's not lying — lying requires knowing the truth. It's autocomplete with a great vocabulary. Treat it accordingly.