Hallucination

A hallucination in AI is when a model confidently generates information that is factually incorrect, nonsensical, or completely made up, presenting it as if it were true. Think of it as the AI being a very convincing liar—it doesn't know it's wrong, so it doesn't hesitate or show doubt. This happens because large language models are prediction engines, not knowledge databases. They work by calculating the most statistically probable next word in a sequence based on their training data, not by retrieving verified facts. When a model lacks the correct data or finds a gap in its training, it doesn't stop to ask for help; it simply invents a plausible-sounding continuation. For example, if you ask an AI for a biography of a fictional scientist you just made up, a model prone to hallucination might provide a detailed birthplace, list of academic achievements, and even a publication history—all of it entirely fabricated. I've seen this happen when asking for a summary of a non-existent research paper; the model gave me a full abstract with fake author names and a journal citation that looked incredibly real. This term is often confused with a simple mistake or a bug. A mistake is when the model has the right information but applies it incorrectly. A hallucination is a deeper fabrication where the information itself has no basis in reality. Understanding hallucination matters because it's the single biggest barrier to trusting AI output. If you're using AI to draft an email, a hallucination might be harmless. If you're using it to research a medical condition or a legal precedent, it's dangerous. You must verify any factual claim an AI makes against a trusted source, treating the model less like an oracle and more like a highly imaginative, overconfident intern.