Glossary · 1 · Foundations: how AI works
Hallucination (AI)
Also known as: Confabulation
A hallucination is output from a generative AI model that sounds plausible but is false, unsupported or invented — for example a nonexistent citation, a wrong specification value or a procedure step that doesn’t exist.
- Beginner
- Technical writers
- Technical marketers
- Technical project managers
In one sentence
AI hallucinations explained: plausible but false output from language models, why it happens and how grounding and review reduce it.
Example
Asked for the torque value of a bolt, a model states “35 Nm” with confidence although the manual specifies 25 Nm — a hallucination with safety impact.
Why it matters on your learning path
- Technical writers: Never publish AI output unchecked. Grounding answers in your approved content and showing sources are the main countermeasures.
- Technical marketers: Hallucinated product claims can create legal exposure. Fact-check every number, feature and comparison.
- Technical project managers: Treat hallucination rate as a quality metric and plan evaluation before launch.
Why it happens
A language model predicts likely text; it has no built-in sense of truth. Missing knowledge, ambiguous prompts and pressure to answer all raise the risk. Grounding, retrieval-augmented generation and the instruction to say “I don’t know” lower it but don’t eliminate it.