Hallucination
Also known as: confabulation
A hallucination is when an AI produces a response that sounds plausible and confident but is factually wrong or entirely made up.
Short answer
A hallucination is when an AI confidently states something false or made up. It happens because the model predicts plausible text rather than retrieving verified facts.
- Applies to:
- Any factual, numerical, or cited output from an AI tool.
- Keep in mind:
- Grounding the model in real source material reduces but never fully eliminates hallucinations.
In more detail
Because a language model generates likely-sounding text rather than retrieving verified facts, it can invent details — fake citations, wrong figures, non-existent features — while sounding certain. Hallucinations are more likely when you ask about niche facts, recent events, or specifics the model was never reliably trained on. Giving the model the source material to work from, and asking it to say when it is unsure, reduces (but does not eliminate) the risk.
Example
Ask an AI for 'three studies proving X' and it may return three real-looking citations — with plausible authors and journals — that don't actually exist. Always check citations against the real source.
Why it matters
Hallucination is the main reason to treat AI output as a draft: anything factual, numerical, or cited must be verified before you rely on or publish it.
Prompts that use this
Sources
- 1.Hallucination (artificial intelligence) — Wikipedia
Wikipedia · verified 2026-07-26
Supports: Definition of AI hallucination as confident but false or fabricated output.
Try a prompt for yourself
Open any prompt and run it in the IterPrompt playground.