- AI rarely invents business facts out of nothing.
- Usually it repeats wrong or outdated details from portals and directories.
- Fix the sources and, over time, you fix the answer.
What is an AI hallucination?
An AI hallucination is a statement a language model makes with full confidence even though it is wrong. With general knowledge this happens because the model fills gaps with likely words. With business data the cause is almost always different: the web contradicts itself and the AI picks the wrong version.
For your business this means the AI quotes opening hours from two years ago, a price that never existed, or says dogs are not allowed although you welcome them. To the customer it sounds just as credible as the right answer.
Where AI gets its facts about you
ChatGPT with web search and Google's AI Mode read several pages for every question and summarise them into one answer. Your website is just one source among many.
- Booking and review platforms
- Maps such as Google, Apple and Bing
- Business directories and regional sites
- Articles, blogs and comparison sites
If a detail appears the same way on many pages, the AI repeats it. If pages contradict each other, the most common version often wins. If your website is the only source with the right value, it loses to five portals with the old one.
The most common wrong details
The wrong details are almost always the ones customers ask about before they book or visit:
- Opening hours and rest days, especially after a season change
- Phone numbers after switching provider
- Prices, for example parking, extras or memberships
- Facilities and rules such as pets, accessibility or a pool
- Addresses after a move
How to fix the cause
You cannot correct the AI itself. You can correct its sources. There are three steps:
- Find: Which pages show the wrong value? Usually more than you think, and many copy it from each other.
- Correct: Change your own profiles yourself. For portals and directories, request a correction with the old and the right value.
- Follow up: After a few weeks check whether the correction is live and follow up if not.
Once most sources are right, the AI picks up the right value. Systems with web search often adapt quickly, pure model knowledge takes longer.
FAQ
Can I tell ChatGPT directly that something is wrong?
In your own chat yes, but it changes nothing for other users. Only correcting the sources online helps for good.
How long until the AI answers correctly?
With ChatGPT web search and Google's AI Mode often a few days to weeks after the sources are fixed. General model knowledge only changes with new model versions.
Is it enough to keep my website up to date?
Unfortunately not. Your website is one source among many. If most other pages show old values, they win.
Harald FischlManaging Director, maxonlineHarald is Managing Director of maxonline, an SEO and GEO agency from Lower Austria focused on hospitality. He works every day on how businesses are presented correctly in search engines and AI answers. More about Harald