Four NLP classifiers tested on 72,134 news articles — Naive Bayes, Logistic Regression, BiLSTM, and fine-tuned BERT. Logistic Regression (F1 0.971) outperformed the BiLSTM (F1 0.962). BERT topped out at F1 0.992. Built with scikit-learn, TensorFlow, and HuggingFace Transformers on the WELFake dataset.