The rapid proliferation of fake news on social media has raised significant concerns about misinformation and its influence on public perception, political discourse, and social stability. This paper presents a machine learning-based approach to fake news detection, leveraging text-based, sentiment, and social context features to improve classification accuracy. We employ Logistic Regression, Random Forest, and Support Vector Machines (SVM) to classify news articles as real or fake, with a comparative analysis of their performance. However, the reliance on classic machine learning models limits the system's ability to capture complex linguistic patterns, which could be addressed by state-of-the-art deep learning techniques like BERT or Transformers. Additionally, we implement a Streamlit-based web application, providing an intuitive platform for real-time fake news detection. Our study highlights key challenges such as data scarcity, evolving misinformation patterns, and psychological biases that contribute to the spread of fake news. The system’s focus on English text and textual features limits its ability to detect misinformation in regional languages or multimodal formats like images, videos, and deepfakes. Furthermore, we propose future enhancements, including the integration of fact-checking APIs, advanced deep learning models like BERT, and multilingual support, to improve detection efficiency. The findings of this study emphasize the importance of AI-driven solutions in combating misinformation and ensuring the credibility of digital media.