Hallucination
Hallucination
A phenomenon in which AI generates false information that sounds completely plausible and factual
In Simple Terms
It's when a generative AI gives you completely made-up information as if it were the right answer. For example, it might describe a fictional event as historical fact, or recommend books and papers that simply don't exist. The AI isn't doing it on purpose — it's a mistake that happens because the system is built to generate an answer even when it doesn't actually know one.
Behind the Name
"Hallucination" literally means seeing or hearing something that isn't there. The term was borrowed because AI confidently producing information that doesn't exist looks a lot like someone experiencing a hallucination — vividly convinced of something that has no basis in reality.
Take a Closer Look!
Hallucination is a phenomenon in which AI — especially large language models — generates incorrect information that sounds completely plausible.
Because the AI delivers it confidently and fluently, as if it were true, there's a real risk that people will take it at face value.
The thing is, most generative AI doesn't understand language the way humans do.
It works by taking the massive amount of data it was trained on and repeatedly calculating what word or phrase is most likely to come next — then stringing those pieces together into a coherent-sounding response. That means it isn't checking whether the content is actually true the way a person would, which is why it can confidently state something even when the information is vague or flat-out wrong.
For example, when asked about something it doesn't know, an AI may not be able to simply say "I don't know." Instead, it tries to construct an answer that fits the context — and ends up making something up entirely.
There are approaches to reduce hallucination, such as Retrieval-Augmented Generation (RAG), which grounds the AI's responses in external, trusted sources. But eliminating it completely is still very difficult.
For that reason, AI output is best treated with some skepticism — cross-checked against reliable sources when accuracy matters.