Ontology-Grounded Retrieval-Augmented Generation. AI retrieval using formal ontologies, not just vector similarity. Schema must define: what entities ARE and how they RELATE (not just that they exist). Build through ontological relationships (isDefinedBy, memberOf, specialty). Verify using L1a gate: check relationship completeness, semantic path traversal. OG-RAG systems preferentially retrieve entities with strong ontological grounding. Emerging paradigm for specialized domains. Schema without relationships fails OG-RAG. Critical for vocabulary sovereignty in technical/specialized fields. Reference: Sharma et al., EMNLP 2025.