This paper introduces the Extended Semantic Networking (ESN) framework . a novel Retrieval-Augmented Generation (RAG)-based architecture specifically designed for constructing dynamic semantic networks. Unlike traditional RAG systems, which retrieve and inject textual snippets, ESN leverages external documents to expand the semantic structure of a conceptual graph. The framework begins with an LLM-generated network and subsequently extends it with concept-relation pairs extracted from real-world documents, forming a hybrid, evolving knowledge graph. ESN is formalized as a dynamic directed graph featuring traceable relation provenance and adaptive edge weights. ESN represents the first RAG system operating at the level of semantic relations, enabling interpretable, adaptive, and structurally rich knowledge modeling. Keywords: RAG, retrieval-augmented generation, semantic networking, LLM, knowledge graph, generative AI