What exactly is an AI 'hallucination' and why does it happen?

An AI hallucination is when a language model confidently generates information that sounds believable but is completely made up or factually wrong. It happens because these models don't "know" facts the way we do — they predict the next most likely word in a sequence based on patterns from their training data. Think of it like a jazz musician who's memorized thousands of songs and can improvise something that sounds musically correct, but the lyrics might be total nonsense. For example, if you ask an AI to write a biography of a famous scientist, it might invent a childhood story or a quote that never existed, simply because the word patterns fit the style of a biography. I've seen models create fake research papers with realistic-sounding titles and author names that don't exist anywhere. The tricky part is that hallucinations are often mixed with accurate information, making them hard to spot unless you already know the subject well. This is why fact-checking AI output is so important — the model itself can't tell you when it's guessing. A useful tip: if you're asking about a specific fact, book, or person, try asking the AI to cite its source. If it can't provide one, or gives a vague answer, treat the information as unverified. For a deeper dive, see our guide on [troubleshooting AI prompts that don't work as expected](/blog/chatgpt-prompts-not-working). **Related**: How can I tell if AI-generated text is accurate? | Do all AI models hallucinate at the same rate?