From user-generated content platforms to shopping portals, an important feature of websites is to identify which contents or items are the most popular and list them on the front page — assuming these contents are also the most interesting. In this paper, we question this assumption by studying data from a major news aggregator website where stories are contributed by users, and by focusing on the consequences of social influence. We first observe that the popularity of contents is correlated with the visibility of their authors, and can thus be explained by social influence. We then provide evidence of a feedback loop between the visibility of contributors and the popularity of their contributions, which increases the importance of social influence over time. Finally we quantify the effect of social influence in making each story popular and show that this effect tends to favor less interesting stories.