PIPELINEGlossary · AI and automation
What is an AI hallucination?
A hallucination is an answer that a language model formulates convincingly but that is factually incorrect or made up. The reason is how it works: the model predicts the most likely next word; it doesn't check facts.
01 — Context
Hallucination in detail
Hallucinations typically occur where the training material had little or contradictory information — with niche knowledge, current events, specific numbers, or proper names. Particularly tricky is that the tone stays identical: a made-up answer sounds just as confident as a correct one.
In business use, you can greatly limit the risk. Three measures work: have the model answer only from verified content, explicitly allow it to say “I don't know,” and for anything with legal or financial consequences, include human oversight.
02 — Practice
What this means for small businesses
For customer contact, this is the critical configuration question. A bot that escalates when uncertain is useful; one that invents opening hours or prices does more harm than it saves work.
03 — Questions
Frequently asked questions
Can you eliminate it completely?
Not completely, but largely in practice. With source binding and permission to say nothing, the rate drops to a level that's viable for customer contact.
How do you recognize a hallucination?
Most often by invented details: numbers, names, deadlines, or references that don't exist. That's why source citation in the bot is so valuable — it's verifiable.
04 — Continue
Related terms and services
- RAG (Retrieval Augmented Generation)
- Prompt
- Automation
- Language model (LLM)
- Knowledge base
- EU AI Act
- Core Web Vitals
- AI Overview
- Webhook
- All terms
Related services
Next step
We'll look at your starting position.
Tell us what's eating up time today. That becomes an assessment with effort and benefit — and only then an offer.