This document describes a project that uses an LSTM recurrent neural network to detect fake news. It trains the LSTM model on a dataset of past news labeled as genuine or fake. The news texts are converted to TF-IDF vectors using n-grams before training. The trained model achieves 69.49% accuracy on the test data at predicting whether a new piece of news text is genuine or fake. Screenshots demonstrate data preprocessing, model training and evaluation, and testing the model on new news texts.