The spread of fake news and misinformation has become a global challenge, undermining public trust, causing political polarization, and facilitating the dissemination of harmful ideologies. This study explores the use of advanced AI models, specifically transformers such as BERT and GPT, for the automatic detection of fake news. Leveraging natural language processing (NLP) techniques like Named Entity Recognition (NER), Sentiment Analysis, and Topic Modeling, we aim to identify patterns unique to misinformation. Our model demonstrates high accuracy in experimental trials on benchmark datasets, highlighting the potential of AI to combat disinformation and improve media literacy.