Artificial intelligence hallucination occurs when an AI model produces output that is plausable but incorrect, often resulting from training data limits.