What is Hallucination?

A hallucination is an LLM output that is fluent, confident, and wrong — a fabricated fact, a misattributed quote, an invented statistic, or a product feature that doesn't exist. For brands, hallucinations are a discoverability and reputation risk: a competitor's positioning can be attributed to you, a discontinued product can be recommended, or a real crisis can be paired with an invented cause. The remedy isn't lawyering the model — it's making the correct answer easier to retrieve and ground against than the wrong one. That means dense, clean earned media, unambiguous entity signals (Organization schema, sameAs, Wikidata where possible), an on-site glossary that defines your terms in your own words, and named-expert quotes the model can lift verbatim. Crackle PR audits every client's hallucination surface quarterly — if ChatGPT or Perplexity is making things up about a brand, the fix is a citation-density and entity-resolution problem, not a legal one.