Let's stop saying 'semantic web' (www.bobdc.com)

The author argues that “semantic web” is a misleading label for technology built around RDF, URIs and inferencing (OWL) because the original vision—a globally interconnected web of machine-readable RDF datasets—never materialized. While core pieces (HTTP, URLs, RDF, SPARQL endpoints) exist, public linked-data is sparse: a few standout SPARQL services like Wikidata and DBpedia power much of the community, but most organizations publish JSON APIs (rarely JSON‑LD) without persistent URIs or interlinked triples. Static RDF files and FOAF demos exist, and standards like SKOS and schema.org have helped interoperability, but the hoped-for public “semantic web” mesh did not emerge. That apparent failure was actually a successful pivot: RDF technologies and related tooling found real traction inside enterprises as internal knowledge graphs, triplestores, and SPARQL-backed data layers—what many now call “enterprise knowledge graphs” or “semantic enterprise standards.” For AI/ML this matters because knowledge graphs are increasingly paired with LLMs and vector embeddings to add structured context, provenance, and to reduce hallucination. In short, the tech succeeded but in different form: prefer talking about RDF technologies, linked data practices, and enterprise knowledge graphs rather than the romanticized global “semantic web.”