What is a Hallucination?

Hallucination is the generation of factually incorrect, fabricated, or unverifiable information by a large language model (LLM) presented as though it were true. Hallucination occurs when an LLM lacks sufficient grounding data, encounters ambiguous entity signals, or cannot corroborate a claim against trusted sources—causing the model to fill gaps with plausible but false content. In AI search optimization, hallucination represents the failure mode that structured identity, knowledge graph verification, and answer-first content are designed to prevent. When a business provides clear, machine-readable entity data through canonical identifiers, validated schema, and consistent fact files, LLMs can retrieve and cite verified information instead of generating approximations. Hallucination rate measures how frequently an LLM produces incorrect statements about a specific entity and serves as an inverse indicator of AI search optimization effectiveness.