Back to Glossary
Concept

Hallucination

When an AI generates information that sounds convincing but is factually incorrect.

Share

Definition

Hallucination is when an AI model generates information that sounds confident and plausible but is actually false, made up, or inaccurate. This happens because AI models are trained to produce likely-sounding text, not to verify facts. They can fabricate statistics, cite non-existent research papers, invent historical events, or confidently give wrong answers to factual questions.

Hallucinations are one of the biggest challenges in AI deployment. They're particularly dangerous in high-stakes domains like healthcare, legal, and finance where incorrect information can have serious consequences. The best defense is combining AI with retrieval systems (RAG), adding explicit "don't make things up" rules to your prompts, and always verifying critical information.

Examples

1

An AI citing a research paper that doesn't exist — complete with a plausible-sounding author, journal, and date

2

An AI confidently stating incorrect statistics about a company's revenue when asked for financial analysis

Related Terms

Frequently Asked Questions

How do I prevent hallucinations?
Three main strategies: 1) Add rules like "If you don't know, say so" to your prompts, 2) Use RAG to give the AI real data to reference, 3) Always verify critical facts from AI outputs independently.
Do all AI models hallucinate?
Yes. All current LLMs can hallucinate. Some models are better than others at acknowledging uncertainty, but none are immune. Always treat AI outputs as drafts that need human review for factual accuracy.

Build prompts using this concept

Explore our prompt library and put hallucination into practice with ready-to-use templates.

Build prompts using this concept