AI assistants generate fluent, confident text, and sometimes that text is simply wrong: made-up hours, a phone number that is a digit off, a service you do not offer, a location you left years ago. This is called a hallucination. The model is not lying on purpose; it is filling a gap with a plausible guess when it lacks solid information.
For a business, a hallucination is a real risk. If an assistant tells a customer you close at 5 when you close at 7, or sends them to an old address, you lose the visit and never know why. The customer blames you, not the AI. The defense is not to argue with the model; it is to remove the gap it was guessing into.
You reduce hallucinations about your business the same way you earn citations: consistent facts everywhere (NAP), schema markup that states your details in machine-readable form, an accurate Google profile, and clear pages the crawler can read. Give the machine the truth, structured and consistent, and it has far less reason to invent.
A customer asks ChatGPT if your shop is open Sundays and it confidently says yes, when you are closed. That is a hallucination, usually caused by thin or conflicting information online about your actual hours.
Go deeper: How to Show Up in ChatGPT and Perplexity Results (2026).
This is part of our AI search (GEO) work.