Hallucination (AI)
What is Hallucination (AI)?
Hallucination is when a language model generates a confident-sounding but factually incorrect statement — inventing an order number, a policy that doesn't exist, or a wrong answer presented with the same fluency as a correct one. It's a fundamental risk of how language models generate text, and CPaaS voice or chat agents mitigate it with retrieval-augmented generation (RAG) to ground answers in real data and guardrails that catch an unverified claim before it reaches a customer.
More detail
Hallucination happens because a language model generates the statistically likely next words based on its training, not because it looks up and verifies facts the way a database query would.
RAG reduces hallucination by grounding a response in retrieved source documents, but doesn't eliminate it entirely — a guardrail or eval layer checking claims against a source of truth is still a common additional safeguard.
Frequently asked
- Why do language models hallucinate instead of just saying 'I don't know'?
- A model generates text by predicting statistically likely words based on its training, not by checking facts against a verified source, so it can produce a fluent, confident-sounding answer even when it has no real basis for it.
- Does RAG completely prevent hallucination?
- No — it substantially reduces the risk by grounding responses in retrieved real documents, but a model can still misinterpret or overstate what a retrieved passage actually says, which is why guardrails and evals remain an important additional check.
See also
Build it on Orbit
Voice, messaging, email, video, and AI agents on one platform and one pay-as-you-go bill. Start free — no credit card required.