This article presents a novel approach to analyzing the credibility of news reports using semantic networking and large language model (LLM) technologies. The proposed methodology enables effective identification of fake news by constructing semantic networks of concepts that reveal causal and associative connections between informational units. Key aspects of this study include the mathematical formalization of network construction, evaluating the credibility of connections, identifying anomalous links, and determining the overall reliability of news reports. The use of LLMs in text analysis enables precise identification of key concepts and their relationships, enhancing the reliability of results. This practical methodology opens up new possibilities for automated credibility monitoring and holds significant value amid modern information challenges. This research contributes to the advancement of information reliability assessment technologies and can be applied to improve systems for detecting manipulation and fake news.