Skip to content

Concept·Last reviewed 2026-08-18

Hallucination

Also known as AI hallucination, model hallucination, confabulation.

What is Hallucination?

A hallucination is a confident, fluent statement from an AI model that is not true, delivered with the same certainty as a correct one. A hallucination is not a malfunction in the ordinary sense, because the model is doing what it was built to do and producing plausible text, which is why hallucinations read as authoritative rather than obviously broken.

Why are hallucinations dangerous for a brand?

Hallucinations are dangerous for a brand because the damaging case is the plausible one, not the absurd one. A wrong price, a discontinued product or a policy the company never had all sound entirely reasonable to the person reading them.

A hallucination also spreads without leaving a trace on the brand. The customer receives the wrong statement inside an assistant, forms a decision on it, and never visits the site where it could have been corrected.

Questions this answers

  • What happens if AI makes something up about my company?
  • Can AI give wrong information about a business?
  • How bad are AI hallucinations for brands?

Do people notice hallucinations?

Most people do not check, so a hallucination usually goes unchallenged. KPMG and the University of Melbourne found in 2025 that 66% of people rely on AI output without verifying whether it is accurate, across 47 countries.

That is what makes a hallucination about a brand costly rather than merely wrong. If two thirds of readers accept the statement as given, the error functions as fact for most of the audience that encounters it.

  • 66% of people rely on AI output without checking accuracy, across 47 countries [1]

Questions this answers

  • Do people fact-check AI?
  • How many people believe AI answers?
  • Do users verify what ChatGPT tells them?

Key takeaways

  • A hallucination is a confident, fluent statement from an AI model that is not true, delivered with the same certainty as a correct one.
  • The damaging hallucination for a brand is the plausible one: a wrong price, a discontinued product, a policy that never existed.
  • A hallucination usually goes unchecked: KPMG and the University of Melbourne found 66% of people rely on AI output without verifying it.
  • A hallucination reaches the customer without ever touching the brand’s own site or analytics.

Citations

  1. [1] Trust, attitudes and use of artificial intelligence. KPMG and University of Melbourne research, accessed 2026-08-18

Changelog

  1. 2026-08-18 · v1.0 · created. Entity created

Machine-readable chunks for this entity: /wiki/hallucination/chunks.json. Corrections to ignite@dhabiai.ae.