AI Systems

Hallucination, explained

A hallucination is model output that sounds plausible but is false, unsupported, or not grounded in the provided context.
Back to Glossary

Part of the Semantic Notion technical glossary.

Short Definition

A hallucination is model output that sounds plausible but is false, unsupported, or not grounded in the provided context.

Intuition

The model can produce fluent text even when it does not actually know or verify the answer.

Technical Definition

Hallucination refers to generation errors where output diverges from facts, source material, task constraints, or verifiable evidence.

Example

A chatbot invents a legal citation that does not exist.

Common Misunderstandings

Hallucination is not only a retrieval problem.
Confidence in wording does not imply factual accuracy.

Same Category

More from AI Systems

Active ParametersActive parameters are the subset of a model's parameters used for a specific input, token, or inference step.AgentAn AI agent is a system that uses a model to choose actions, call tools, and make progress toward a goal over multiple steps.Fine-TuningFine-tuning adapts a pretrained model by continuing training on task-specific or domain-specific data.GuardrailA guardrail is a system control that constrains or monitors AI behavior to reduce unsafe, invalid, or undesired outputs.InferenceInference is the process of running a trained model on new inputs to produce outputs.Instruction TuningInstruction tuning trains a model to follow natural-language instructions and produce useful responses.

Related Reading

Essays that use this concept

AI EvaluationHow to Evaluate AI Systems in ProductionVoice AI SystemsCommon Failure Modes in Voice AI Agents

Start here

Need the broader concept map?

Return to the glossary index to move from this definition into adjacent AI, ML, mathematics, and computer science terms.

Browse Glossary