An AI hallucination is generated output that is unsupported, incorrect, fabricated, or inconsistent with available evidence.
The term describes an output failure rather than intent to deceive, and it is especially important when using generative AI for documentation, research, support, or decision support.