The rise of social media has made it increasingly challenging to differentiate between genuine and fake news. The prospect of detecting misleading news is now more crucial than ever due to the growing dependence on social media, particularly on smartphones, for accessing news. This paper evaluates the performance of advanced deep learning models, namely LSTM networks and BERT. Social media is user-friendly, and so more individuals come across false in-formation; this makes it difficult to detect such content using methods based on content analysis that do not consider context. In this paper, a user profile based geometric deep learning model (GY) is proposed that considers content, social graphs, and news dissemination patterns. Here, the model accurately detects fake news within hours of its circulation. This establishes that NLP based methods perform well in the detection of misleading information. Moreover, this research reveals that the DT algorithm attained a higher accuracy than SVM.