Why does ChatGPT sometimes make up fake facts and sound so confident about them?

ChatGPT makes up fake facts because it's designed to predict the next most likely word in a sentence, not to check if that sentence is actually true. It's a text predictor, not a fact-checker. When you ask it a question, it scans its training data for patterns and assembles a response that sounds plausible. If the training data is thin on a topic or the prompt pushes it into obscure territory, the model will fill the gaps with invented details. And it'll do this with complete, unshakable confidence. I've seen it invent book titles, academic citations, and even historical events that never happened. The industry calls this 'hallucination,' but that's a bit misleading. The AI isn't hallucinating in a human sense. It's doing exactly what it was built to do: generate coherent text. The problem is that coherence and factual accuracy are two completely different things. A good example: if you ask ChatGPT for a biography of a minor 18th-century painter, it might invent a birth year, a famous patron, and a dramatic death — all wrapped in polished prose. It's not lying. It just doesn't know what it doesn't know. The fix isn't simple. You can reduce hallucinations by giving the AI very specific source material and asking it to stick to that text. But the risk never fully goes away. This is why using AI for research without verifying the output is dangerous. For a deeper dive, see our guide on troubleshooting why your prompts produce bad results. **Related**: How can I tell if AI-generated text is accurate? | What's the difference between an AI hallucination and a simple mistake?