<div>This paper develops a theoretical framework to examine how a social media platform allocates attention through recommendation algorithms and how this in turn shapes content creation and consumption. Creators and viewers, as the two sides of the algorithm, fall into different categories based on interest. Creators are also heterogeneous in ability. We show that a platform, to maximize advertisement revenue, optimally filters out low-ability creators, restricts the reach of medium-ability creators to relevant audiences only, and propagates viral content for high-ability ones at the expense of relevance. The attention a creator receives grows disproportionally in his ability and the popularity of his category. We show the source of the inefficiencies of the algorithm by contrasting it with a welfare-maximizing benchmark. We additionally study the effect of monetary transfers in the algorithm. Our framework offers insights into content production and matching in digital markets, giving rise to potential regulatory interventions.</div>