SantageAI Glossary › AI Hallucination
AI Glossary

What is AI Hallucination?

AI hallucination occurs when an AI model generates information that appears plausible but is factually incorrect, fabricated, or unsupported by its training data.

What is the core idea behind AI hallucinations?

The model does not know what it does not know.

How do AI hallucinations differ from related concepts?

ConceptDifference
Hallucination vs ErrorErrors are mistakes. Hallucinations are fabrications presented as fact
Hallucination vs BiasBias is systematic skew. Hallucination is invented information
Hallucination vs UncertaintyUncertainty is acknowledged. Hallucination is false confidence

How do AI hallucinations work?

What are the limitations of AI hallucinations?

Why are AI hallucinations important?

Hallucinations are a fundamental limitation of current language models and a key reason why AI outputs should always be verified before use in consequential decisions.

How are AI hallucinations used in practice?

Hallucinations have been observed in every major language model. Mitigation strategies include retrieval-augmented generation (RAG), grounding in verified data sources, and training models to express uncertainty.

Frequently Asked Questions

Why do AI models hallucinate?
Language models generate text by predicting the most likely next token based on patterns in training data. They do not have a built-in mechanism for distinguishing fact from fiction or for knowing when they are uncertain.
Can hallucinations be fully prevented?
Not with current technology. Hallucinations can be reduced through techniques like RAG and grounding, but they remain a fundamental characteristic of how language models generate text.