What is AI hallucination?
An AI hallucination is an answer from an AI system that sounds confident and plausible but is false or not supported by any source. It is a known limitation of language models.
Why it matters
In customer service, a wrong answer can cost money and trust. A customer who is told about a discount or policy that does not exist may act on it and then complain. Hallucinations are hard to spot because they are written as fluently as correct answers. Any team that uses AI with customers needs a plan to reduce and catch them.
How it works
A language model writes by predicting likely words, not by checking facts. When it lacks the information it needs, it can fill the gap with something that merely sounds right. Grounding answers in retrieved documents, instructing the model to say when it does not know, and having people review samples of conversations all lower the rate. None of these removes the problem entirely.
Example
A customer asks an AI agent whether a subscription can be paused. No help article covers pausing. An ungrounded model might reply that plans can be paused for up to three months, which is invented. A grounded agent replies that it cannot find this in the help content and offers to pass the question to the support team.
