The subject disclosure relates to leveraging input from media consumers of media content with no or unstructured metadata in order to provide accurate and relevant structured metadata for the media content. Thus, media content uploaded to a content server can be associated with a set of predetermined metadata tags that are intended to be relevant to the media content and to accurately describe the media content. These tags can be implemented as key-value pairs, in which either the keys or the values for a given key can be provided by users (e.g., consumers of the media content), and any such input can be assigned a confidence score in order to determine the “correct” value for a key.