Trust, but verify.
The most expensive ChatGPT mistake isn't using it — it's believing it at the wrong moment. Confidence is not accuracy. Here's the quick verification habit that lets you move fast anyway.
01 Why it gets things wrong (and why it sounds so sure)
ChatGPT generates the most plausible next words given everything it has seen. Usually, plausible and true overlap. Sometimes they don't — and the model has no internal alarm bell for the difference. That's why wrong answers arrive in the same confident tone as right ones. This isn't a scandal; it's a property of the tool, like a chainsaw's kickback. You don't refuse to use the saw — you learn where to stand.
02 The risk triage: what to check
| Claim type | Risk | What to do |
|---|---|---|
| Brainstorms, drafts, rewrites, explanations of concepts | Low | Use freely — you're the editor anyway |
| Numbers, dates, prices, statistics, names | Medium | Verify before repeating to anyone |
| Laws, taxes, medical, contracts, anything you'll be quoted on | High | Verify with a primary source, every time |
03 The 60-second workflow
Suspiciously specific details in answers-from-memory — exact statistics, precise quotes, perfect-sounding citations — deserve the most suspicion, not the least. Fabricated details are usually crisp; real memory is usually hedged.
04 Make errors cheap
The deeper strategy: arrange your workflow so ChatGPT's mistakes get caught by design. Use it for drafts a human reviews, plans you sanity-check, options you choose between. Never wire its raw output straight into anything that matters without a checkpoint. Fast and supervised beats slow and fearful — and both beat fast and blind.
Ask ChatGPT a factual question from your own field — something you know cold. Watch how it does. Then ask it something adjacent that you can't verify yourself, and notice how differently you should treat that answer.
Open ChatGPT →This week's challenge
This week, catch one confident wrong answer. Ask about topics you know deeply until you find one. Screenshot it and keep it somewhere you'll see it — a permanent, personal reminder of what this tool is and isn't. Everyone who uses AI well has one of these.