What it means
A hallucination is when an AI model produces information that sounds plausible but is false, unsupported or invented, such as a non-existent source, a wrong date or a fabricated quote. It happens because language models generate likely text rather than checking facts.
Real-world example
A researcher asks an AI assistant to list publications on the history of a museum's Egyptian collection. The assistant returns eight references with authors, titles and years. Three of them do not exist. They look entirely realistic, which is exactly why they are dangerous.
Why it matters for AI adoption
Hallucinations are a fundamental property of current generative AI, not a bug that will simply be fixed. Techniques such as RAG, source citations, constrained outputs and human review reduce the risk but do not eliminate it. Any use case where errors could harm people, decisions or trust needs safeguards designed for this.
Questions to ask
- How does the system show where its information comes from, and can users verify it?
- What happens if a wrong answer gets through, and who is responsible for catching it?