Within the context of addressing misinformation on social media platforms, the identification of fake news emerges as a crucial obstacle. Nowadays, organizations across many sectors are struggling to find efficient ways to detect online phoney news. Identifying false material online can be mentally stimulating, often crafted to mislead visitors. Deep learning enables algorithms for fake news detection more accurately than other machine learning. These algorithms are characterized by high accuracy and the use of neural networks that allow them to effectively detect fake news. This work outlines an approach that improves upon previous works utilizing the Truth Seeker 2023 dataset. Another part of the methodology is to use natural language processing algorithms to clean the tweets and use neural network architectures to identify subtle signs of false leads. The Multilayer Perceptron (MLP) model consist of densely connected hidden layers with specific activation functions. To convert textual information into numerical characteristics, such as Count Vectorizer and Term Frequency-Inverse Document Frequency TF-IDF, the most proficient text vectorization methods were adopted. Our model achieves an exceptional accuracy above 99% by conducting thorough experimentation and optimization, which is a significant improvement compared to previous studies’ highest accuracy of 96%. We demonstrate our method’s unmatched accuracy in identifying fake news reports by utilizing deep learning and the comprehensive features of the Truth Seeker 2023 dataset. Our research sets a new standard for spotting fake news. It shows how advanced technologies may reinforce social media platforms against misleading information, creating a more dependable digital environment for people worldwide.