Constructing ontologically coherent subsumption hierarchies for perdurant entities (e.g., events, processes) is a complex and labor-intensive task. While prior work has focused on endurants, automated support for perdurants remains limited. This paper examines whether large language models (LLMs), specifically GPT-4, can assign four ontologically grounded meta-properties to perdurant types. We design a prompting framework incorporating property definitions, domain context, and perspective, and evaluate it against a gold-standard dataset derived from the MAVEN taxonomy. Results show high agreement with human annotations under well-specified prompts, indicating GPT-4’s potential to reduce manual effort in perdurant classification. Our contribution formalizes the conditions enabling reliable LLM-based meta-property assignment for scalable ontology construction.