Definition of AI hallucination as false fact

Summary

In AI, a hallucination (also called artificial hallucination) is an AI-generated response that includes false or misleading information while presenting it as if it were true. The term is loosely compared to human hallucinations, which involve perceiving things that are not actually there, but in AI the issue is about fabricated or incorrect facts appearing in the output. This can happen especially with chatbots built on large language models (LLMs), which may produce plausible-sounding random falsehoods—for example, inventing citations or details that were never provided or verified. Detecting and mitigating these errors is difficult, particularly for high-stakes uses such as medical diagnostics, chip design, and logistics, where incorrect “facts” can cause real harm. The label “AI hallucination” is also criticized by some researchers because it anthropomorphizes the system and can be vague about the underlying causes of the errors.

Key Points

• 1

  AI hallucination is defined as false or misleading information generated by an AI and presented as fact.

• 2

  Large language model chatbots can embed plausible-sounding falsehoods (e.g., fabricated citations), making detection and mitigation challenging in high-stakes settings.