<p>Geological knowledge is fundamentally interpretive, yet the semantic infrastructure meant to formalize it remains fragmented, unevenly preserved, and insufficiently characterized. This survey applies a stratified evaluation framework to 28 geoscientific semantic artefacts spanning axiomatized OWL ontologies, schema-based data models, and literature-only resources. Available ontologies are evaluated through two complementary axes: intrinsic quality, combining FAIR-conformance auditing with structural and hierarchical metrics, and ontological interoperability, analysed through a layered interoperability pyramid. Schema-based data models are examined through a dedicated schema-level assessment protocol, while orphaned resources are analysed through a proxy evaluation procedure, revealing that 64.3% are no longer retrievable as machine-accessible artefacts. Among surviving ontologies, structural complexity varies by orders of magnitude; foundational resources achieve reuse potential through controlled expressivity and architectural restraint, while schema-based standards ensure operational interoperability at the cost of implicit and underspecified semantics. A controlled Retrieval-Augmented Generation (RAG) experiment is introduced as a diagnostic comparison with the human-led evaluation, showing that retrieval-grounded synthesis can recover broad documentary patterns but tends to conflate schema-level organization with formal ontological commitment and cannot reproduce structural asymmetries that require artefact-level inspection. Collectively, these findings expose a geoscientific semantic ecosystem that is simultaneously rich and fragile: conceptually ambitious, institutionally uneven, only partially interoperable, and still missing an explicit formalization of the observational-to-interpretative reasoning at the epistemic core of geology itself.</p>