This paper addresses the challenge of detecting fake news using machine learning algorithms due to the limitations of manual detection. It evaluates the performance of eight machine learning algorithms (regression, support vector classification, multi-layer perceptron, gaussian and multinomial naive Bayes, random forests, decision trees, and convolutional neural networks) across three datasets, finding that a hundred-dimensional space with TF-IDF vector generation is effective for high accuracy. Convolutional neural networks emerged as the top performer despite longer training times, with pretraining on benchmark datasets offering similar benefits to training on specific data.