What Are AI Hallucinations?

# What Are AI Hallucinations?

You ask ChatGPT for a list of sources on a topic. It gives you five academic papers with authors, titles, journals, and publication years. They look perfect. There's just one problem: three of them don't exist. The authors are real, the journals are real, but those specific papers were never written. The AI invented them — confidently, convincingly, and without any warning.

This is an AI hallucination. And it happens far more often than most people realize.

Defining Hallucination

An AI hallucination occurs when a language model generates output that sounds plausible and confident but is factually incorrect, fabricated, or unsupported by any real information.

The key characteristics: - Confident tone — The AI doesn't hedge or express uncertainty. It states falsehoods with the same conviction as truths. - Plausible structure — The output follows the expected format (citations look like citations, statistics look like statistics). - No source basis — The information wasn't in the training data in that form. The model generated it by pattern-matching.

Hallucination is different from other AI errors: - Bias is when AI reflects prejudices from its training data. The information may be "real" but skewed. - Outdated information is when AI gives correct information that has since changed. It was once true. - Hallucination is when AI generates information that was never true. It is fabricated.