Why AI Makes Things Up, and How to Stop It

Yes, sometimes it does. AI can say something wrong with total confidence. That's real, and being careful about it is smart.

Here's how you cut it down to almost nothing.

1. Give it the source.

   Don't ask it to answer from memory. Paste the document, the numbers, the email, and say “only use what I gave you.” It can't make up what it's reading off your page.

2. Tell it to admit when it doesn't know.

   Add this line: “If you're not sure, say so instead of guessing.” This one line changes everything.

3. Ask it to show its work.

   Try “Show me where this came from” or “What could be wrong about this answer?” It will flag its own weak spots.

4. Have it check its own work.

   For anything detailed or fact based, add this line: “QC your result for accuracy before responding to me.” It reviews its own answer and catches more of its mistakes before you see them.

5. You stay the boss.

   AI is a first draft and a research assistant, not the final word. Check anything that matters before it goes out. Names, numbers, dates, legal or medical claims, every time.