Selecting Diverse, Relevant Content From Multiple Content Feeds

A digital magazine server creates cover pages identifying relevant content items relevant to a user. Candidate feeds of content items are identified from various sources, such as user-defined sections of content items, social networking systems, external content providers, and recommended content. The digital magazine server retrieves content items from the candidate feeds and generates clusters including retrieved content items based at least in part on the content elements contained in the content items. The content items in various clusters are scored, and one or more content items are selected from each cluster. The selected content items are placed in a consolidated feed, which is used to create a cover page describing a digital magazine for presentation to a digital magazine server user.