The AI system process of determining which real-world entity corresponds to an ambiguous reference — resolving ‘Michael Jordan’ into the basketball player vs. the Berkeley professor vs. the actor. Disambiguation relies on structured data, corroboration, and context. Entities that fail disambiguation are omitted from AI responses because systems cannot confidently determine which entity is meant. KGMID assignment is evidence of successful disambiguation. Schema.org markup, comprehensive sameAs linking, and consistent NAP (Name-Address-Phone) across sources are the technical inputs to disambiguation.