A hallucination is when an AI states something false with confidence. For a brand, the stakes are specific: the false thing an engine tells your customers about your price, your product, or your history, in the same certain voice it uses for the truth.
Why it matters
Engines do not flag their own mistakes. A made-up detail and a real one arrive in the same fluent sentence, and a customer has no way to tell them apart.
For brands the failure is not abstract. It is an engine confidently telling a buyer something wrong about you, and it repeats until something corrects the record it is drawing from.
The hard part is judging it honestly. Calling a claim a hallucination when you cannot actually source the truth is its own kind of error.
How Jinn treats it
Fama does not grade a claim as simply true or false. Its verdict vocabulary has seven states, including outdated for what was once true and unverifiable for what no ground truth can settle.
One rule guards against overreach: a claim that reads wrong but carries no quality source is never published as wrong. It lands in a disagreement state instead, and a brand is never scored down on a claim we could not source. Honesty about what we cannot verify is part of the measurement.
Fama catches what the engines say that is provably false, grounded in your verified Brand DNA. See it
Related terms
The stakes are what engines tell your customers.
A confident, wrong answer about your brand repeats until the record is corrected. See how Fama catches what is provably false.