First, the researchers created dictionaries containing the words that fall under each studied category (persuasive, uncertainty, positive and negative emotion, and comparative words). Using Chinese social media platform Sina-Weibo, researchers found 9,631 fake-news posts. Each fake post was matched with a true post that was similar in content length, creator's number of followers, number of likes and comments, and number of references to other sources. This effectively controls for number of views on a post, which cannot be measured directly using this method. The number of reposts for the matched posts were compared. The information richness of each post was coded as 1, 2, or 3 (text only, text + picture, text +video).