Metadata type, not quantity, determines LLM grounding in museum descriptions

Research shows that adding more metadata does not inherently improve LLM grounding for museum artifact descriptions. Instead, the specific type of field provided dictates whether models ground information in fact, omit details, or rely on unsupported inferences, while images can inadvertently increase hallucination via unsupported interpretation.