Hallucinations — why AI is sometimes confidently wrong
Learn why AI sometimes confidently makes things up and how a knowledge base keeps your agent grounded in real facts.
This is one of the most important lessons. Not to scare you — but to make you a smarter user who never ends up in an awkward situation.
What is a "hallucination"
Sometimes AI confidently states a fact that doesn't exist. It invents a book with a real author's name. It gives a non-existent address. It cites a law that was never written.
This is called a hallucination — and it's not a bug in any particular agent. It's a feature of how all AI works.
Remember the previous lesson: AI doesn't look up information, it predicts plausible text. And sometimes plausible doesn't mean correct.
A real-world example
Carlos is the owner of a small law firm in Mexico City. One day he asked the agent, with no additional context: "Write about our firm — what we do, who our clients are."
The agent wrote beautifully. But it listed practice areas that don't exist, invented "10 years of experience," and added a specialization the firm doesn't have.
Why? Because the agent had no real data. It filled in the gaps with what "sounds true" for a law firm.