When AI lies to you (and it will)

AI will sometimes invent facts, quotes, even court cases — with a straight face. This is the most important lesson on this site.

The lesson

What "hallucination" means in plain English: AI predicts plausible words; it doesn't check a fact database. You'll see a confident wrong answer happen live, then learn the 2-source rule: any fact that matters gets verified somewhere AI didn't write.

The picture to keep

The traffic light: 🟢 brainstorms, drafts, explanations you can sanity-check · 🟡 summaries of things you can verify · 🔴 facts, figures, citations, anything you'll act on.

Try it — 60 seconds

Tell me about [a topic you know deeply — your job, your hometown, your hobby]. — Then grade the answer. Find what it got subtly wrong.

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

Watch out

The scariest errors aren't the absurd ones — they're the plausible ones. That's why you test AI on something you already know.