Hallucination
In artificial intelligence, a hallucination occurs when a large language model mistakenly perceives patterns or objects that do not exist. In the context of RAG, an ungrounded answer includes a guess or hallucination unsupported by the source documents.
On this page
Source content checked . English translation published 2026-10-27.
In business automation
A hallucination is content that sounds convincing but is not based on accurate information. In automation, it is especially risky when text becomes a system field or a promise to a customer without being checked.
Examples
These examples illustrate a possible workflow. Adapt the connections and rules to your own systems.
Business workflow example
A bot is asked whether a product is in stock. Instead of letting the model make up an answer, the workflow queries the inventory system and displays the result. If the query fails, it says it cannot verify availability.
Relevant tools
- Chatbase
- A knowledge-based bot still needs its answers checked and its behavior defined for cases where there is not enough information.
What to check
- Adding a source or an instruction not to invent does not eliminate errors. Check that the source actually supports the answer and that missing information remains marked as missing.