Skip to content
Glossary

AI hallucination, explained

A hallucination is when an AI model produces information that looks correct but is false or not supported by the data it received.

Language models generate the most likely text, not necessarily the true one. Without reliable context, they can invent a deadline, a price or a policy.

In customer service, hallucinations are controlled with a reviewed knowledge base (RAG), instructions telling the agent to admit when it does not know, limits on sensitive actions and conversation reviews.

Tracking the rate of unsourced answers and handoffs due to uncertainty helps monitor the risk over time.

Conversations aren't support.They're infrastructure.

Step into the new era of the relationship between companies and customers.

Create agent now