People often discover that an AI chatbot has given a detailed answer that turns out to be wrong. This behaviour is commonly called hallucination, and it is one of the best-known limitations of current generative AI.

Why it happens

Language models are designed to predict plausible next words. They are not looking facts up in a verified database unless they are connected to one. When the model lacks reliable information, it may still produce a smooth answer that sounds right but is not.

Common warning signs

  • Very specific names, dates or citations that you cannot find elsewhere.
  • Links or references that do not exist.
  • Answers that change when you ask the same question again.
  • Confident claims about recent events the tool may not know.

How to protect yourself

Cross-check important claims with reliable sources such as official websites, reputable news outlets or published research. Ask the tool to explain its reasoning or to say when it is unsure, and give it the source text when accuracy matters.

Where the risk is highest

Medical, legal, financial and academic work deserve the most care, because errors there can have real consequences. In those areas, use AI as a starting point and rely on qualified professionals for decisions.

Takeaway

Hallucinations do not make AI tools useless. They mean users should verify what matters and treat fluent language as a feature of the tool, not as evidence of truth.