Why AI Makes Things Up, and How to Reduce It

AI models predict likely words. When they lack the facts, they can still produce a confident, fluent answer that is wrong. This is often called a hallucination.

How to reduce it

• Ground answers in your own documents instead of the model's memory.
• Ask it to cite the source for each claim.
• Let it say "I do not know" when the material is missing.
• Add review for anything important.

Test it

Collect real questions with known answers and check how often the system gets them right. Do this before launch and after every change.

You will not remove every error, but you can make them rare and easy to catch.