This study investigates the relationship between the sentiment of news on social media and readers’ emotional response. We analyzed 1946 teaser headlines (‘New life for a dead language, as more public schools offer Latin’) from the Wall Street Journal’s Facebook page and determined their sentiment with the program VADER. We find that emotional headlines lead to more reactions than neutral headlines. The sentiment scores were systematically related to the emoji the readers attached to the headline (respectively those for ‘like’, ‘love’, ‘haha’, ‘wow’, ‘sad’, and ‘angry’). News with a negative sentiment is related to an increase in negative emotional responses by readers (as expressed in the emoji they use, ‘angry’ and ‘sad’) as compared to neutral and positive news. However, positive news does not result in an increase of positive affective responses (e.g., positive emoji) compared to neutral or negatively colored news. These results are in line with earlier positive-negative asymmetry findings, showing that negative events lead to a higher need for action than positive events. They also underscore the complexity of predicting emotional responses, because we also found some unexpected effects, such as the fact that a ‘haha’ response was given to both positive and negative headlines and that ‘wow’ responses were more prevalent with negative news. This study therefore contributes to our understanding of how people use emoji.