Hallucination
When an AI model generates plausible-sounding but factually incorrect information. In agent contexts, hallucination is mitigated through RAG (grounding responses in your knowledge base), confidence scoring, and citation requirements. Critical in legal, healthcare, and finance agents where accuracy is non-negotiable. In multi-step agents a hallucination is especially dangerous because it *cascades*: the fabricated fact becomes a false premise the agent reasons confidently from at every later step — the grounding-failure form of Compounding Error, and a core reason reliability degrades over long tasks.
Written by Max Zeshut
Founder at Agentmelt · Last updated Sep 9, 2026
Example
A support agent invents a return policy that does not exist. RAG-based grounding prevents this by requiring the agent to cite specific knowledge base articles.
Frequently asked questions
- How do you prevent AI hallucination?
- Common techniques include RAG (grounding answers in real documents), confidence thresholds (declining to answer when uncertain), citation requirements, and human-in-the-loop review for high-stakes responses.