There has been a heated debate over the last few months over the question of how many real human users Twitter has. While all previous work [3], [4], [5], [12], [13], [15] focused on analysis of Twitter accounts and attempts to classify bots, Similarweb takes a different, unique approach.

We apply an explainable machine learning algorithm using datasets of digital activities of panels of users to provide robust estimates of Twitter's average Monetized Daily Active Users (mDAU) for the US over the period of Q3 2021-Q2 2022. Our approach has several advantages that allow us to provide estimates for various user populations beyond what is reported by Twitter or investigated in previous work. For one, our approach allows us to ensure with a significantly higher level of confidence that a user is not a bot using their complete digital footprint beyond just tweets. We can further estimate the additional unauthenticated users that do not hold a Twitter account yet consume content on Twitter and as such can be monetized. We provide a clearer insight into the fact that a relatively small portion of authentic Twitter users is responsible for most of the content published by authentic users. Finally, our study shows that, even under the assumption that a relatively small number of active bots generate content, their overall content contribution is actually significant.