We study how social image concerns affect information sharing patterns between peers. An individual receives a signal ("news") about the state of the world and can either share it with a peer or not. This signal has two attributes: a headline -- e.g., arguing for or against human-induced climate change -- and a veracity status, indicating if the signal is based on facts or made-up. The headline is observable at no cost by everyone, while observing the veracity status is costly and the cost depends on an individual's type. We study the sharing patterns induced by two different types of social image concern: wanting to be perceived as talented, which implies being able to distinguish proper from fake news, and wanting to signal one's worldview. Our model can rationalize the empirical finding that fake news may be shared with a higher propensity than proper news (e.g., Vosoughi et al., 2018). We show that both a veracity and a worldview concern may rationalize this finding, though sharing patterns are empirically distinguishable and welfare implications differ.