New: outcomized services. Pay per result, with a monthly cap you set. See pricing →
AI & LLMs

What is Hallucination?

Hallucination: When an AI model produces confident output that is false or not supported by its sources, such as invented facts, citations, or policy details.

A hallucination is output that sounds plausible but is wrong or made up. Models generate the most likely continuation of text, not verified truth, so when they lack information they can fill the gap with something that fits the pattern: a refund policy that does not exist, a citation to a paper nobody wrote, an order status the system never returned.

Hallucinations are reduced, not eliminated, by grounding the model in retrieved sources, instructing it to say when it does not know, requiring citations, constraining output with schemas, and checking answers against data before they reach a customer. Newer models hallucinate less than older ones, but no production system should assume a zero rate.

Why it matters when you build or buy AI

The question is not "does it hallucinate" but "what happens when it does." Design for it: a support bot should hand off when it cannot cite a source, and an agent that issues refunds should check amounts against the order system before acting. Any workflow where a wrong answer emails a customer or moves money needs a check or a human approval step.

Example

A support bot is asked about a warranty that the help center never mentions. Instead of inventing terms, it says it does not have that information and offers a human handoff.

Is your agent live and unmeasured?A one-week audit: evals, failures, cost per task, permissions. $2,500.

Related terms

More in AI & LLMs

Next step

Book 20 minutes, or send a brief

A former CTO replies. You leave with a written plan and a fixed estimate.

No spam. One reply within one business day. Or skip the form and book a call.

Thank you. Your brief is on its way. Grab a time now so we can talk it through.

Prefer LinkedIn? Connect with Harish.

Book a callFree lead test