---
title: "Hallucination (AI)"
description: "In AI, a hallucination is a confident answer that is not supported by the model's input or by facts: an invented price, a policy that does not exist, a citation"
canonical: https://automataai.hu/en/glossary/hallucination
language: en
publisher: Automata AI
updated: 2026-09-19
---

# What is Hallucination (AI)?

In AI, a hallucination is a confident answer that is not supported by the model's input or by facts: an invented price, a policy that does not exist, a citation that was never written. It is the main risk in customer-facing AI and is controlled by grounding answers in your documents, restricting scope and routing uncertainty to a person.

## Four controls that work

Ground answers in retrieved documents. Restrict the assistant to its domain and let it say 'I don't know'. Validate structured outputs against your systems. Log every conversation and review a sample weekly. Together these bring customer-facing error rates to a level below a busy human agent's.

## FAQ

**Can hallucinations be eliminated completely?**

No, but they can be made rare and harmless: rare through grounding and scope, harmless by never letting the model act on uncertain information without a check.

Related: https://automataai.hu/en/glossary/rag, https://automataai.hu/en/glossary/llm, https://automataai.hu/en/glossary/human-in-the-loop