Hallucination

A hallucination is when an AI states something false as though it were fact. The tricky part is that it does not sound false. Hallucinations tend to be fluent, confident, and specific, which is exactly what makes them easy to miss.

To understand why this happens, remember what a large language model actually does. It predicts likely text. It is not looking anything up in a database of verified facts. Most of the time the most likely text is also true, because true statements are common in what it read. But when it hits a gap in what it knows, it does not stop and say so. It produces the most plausible-sounding continuation, and sometimes that continuation is wrong.

This shows up most with precise details: a specific date, a page number, a citation, a court case, a quote attributed to a real person. The AI may generate a reference that looks perfectly formatted and does not exist.

The good news is that this is manageable once you expect it. Treat AI as a fast, capable first-drafter rather than a source of record. For anything that matters, check its claims against a reliable source, and be especially careful with exact figures and citations. Asking the AI to tell you how confident it is, or to note where it is unsure, can help too. None of this means AI is untrustworthy. It means you stay the editor.