Misinformation is a growing threat in digital environments, especially in fictional media such as films and podcasts, where manipulated content is often subtle yet impactful. This research presents a method based on LLMs and RAG to detect such misinformation. The proposed architecture integrates Large Language Models (LLM) with Retrieval-Augmented Generation (RAG) techniques to construct a dynamic ontology based on verified information, which is subsequently employed to query and authenticate the content of fictional works. This system identifies discrepancies between the verified knowledge base and the information presented in these works. The tool was evaluated using textual data, yielding promising results characterized by high precision and recall, thereby highlighting its potential for detecting misinformation across a variety of fictional media. These findings suggest a robust solution for enhancing content verification in an ever-evolving digital landscape.