In contemporary U.S. culture celebrities compete for attention and publicize their work using social media tools. Twitter is a popular platform that celebrities use to post a variety of content, however little is known about the potency of these different types of content to draw audience attention and participation. In this paper, we outline a scheme for classifying content created by celebrity users on Twitter and analyze the audience engagement to these diverse types of tweets. We find that different types of content produce different levels of audience engagement and that a celebrity’s everyday usage of Twitter (selfies, photos of travels, humour, etc.) produces the most engagement, followed by self-endorsement and commentary about society. But we also find that these patterns vary between celebrities, and audiences are not identical in their response to the content. We likewise determine that there is some other source of unexplained variation, likely resulting from Twitter’s recommendation algorithms or external media coverage. As part of our study, we propose a unique coding scheme which enhances the effectiveness of inter-coder reliability, reducing the time to code.