Social bots are automated agents programmed to act on social media impersonating human behaviour to influence discussions online. This paper aims to contribute to the discussion of how bots can endanger online communication and alter information flows. We resorted to a mixed-method approach based on grounded theory and observational techniques in order to investigate the bots’ activities online during the 2016 municipal elections in Rio de Janeiro. We collected related content on Twitter in this period and detected 3,101 bots. This sample was classified in three categories based on tweeting content: user-generated bot, media spambot, and political bot. Our findings indicate that, although bots work for different political and social purposes, their computational nature claims into service of dominant social groups and economical elites. We conclude that computational propaganda is building a dangerous scenario of widespread automation in which different kinds of algorithms bias social media conversation.