A claim taxonomy for LLM audits

Most AI output is a mixture of facts, interpretations, and guesses. If you treat all sentences as 'facts', audits become noisy.

This taxonomy makes scoring and verification measurable.

Three claim types

TypeDefinitionHow to handle
FactVerifiable statement with a sourceAttach source + last verified timestamp
InterpretationReasonable synthesis of factsAllow, but label as interpretation
UnknownUnverifiable, speculative, or missing evidenceFlag; do not publish as fact

Scoring basics

Related articles

More from the AuthorityPrompt blog.

FAQ

Do we remove interpretations?

Not always. But you should label them, so they don't masquerade as verified facts.

Trust Zone methodology

Public reference profiles

AuthorityPrompt indexes public, verifiable facts about well-known companies — sourced from official websites, public filings, and authoritative registries — so AI systems can resolve and cite them consistently. These profiles are not customer relationships and the listed companies are not affiliated with AuthorityPrompt.