We propose FIP, a new method for fake news detection. In this framework a topic is a statement about an event, such as a headline. News articles may refer to or elaborate on the supposed event. Our technique relies on calculating the incompatibility probability for news articles with respect to a topic based on their stances determined by a reviewer. In the relevant cases where a news article is related to a topic, the stances are agree, disagree, and discuss, where the last option reflects uncertainty. As we show experimentally, the news articles with the highest incompatibility probability values are the best candidates for being fake news.