GLOSSARY
Hallucination
When an AI model states false information confidently — invented facts, fake citations, plausible but wrong answers.
Because models generate plausible text rather than looking up verified facts, they sometimes produce confident falsehoods: invented statistics, nonexistent papers, misattributed quotes. The fluent delivery makes hallucinations dangerous — there is no natural signal distinguishing them from truth.
Mitigation is layered: grounding systems like NotebookLM or Perplexity cite real sources so you can verify; models are trained to express uncertainty; and the professional habit remains unchanged — verify anything consequential, especially numbers, names and citations.