The application of AI commenting robots has sparked a new wave of online interactions and triggered perceived AI-generated trolling. Given the spread comments by social bots, it is critical to gain a better understanding of their perceived trolling comments. This study examines the trolling tactics and textual features of perceived social bot trolling in China, focusing on Sina Weibo’s CommentRobot. Through thematic content analysis of 716 interactions of perceived CommentRobot trolling from RobertVictimsAlliance, we identified frequent use of trolling tactics that resemble those used by human beings, and most commonly these involved derailing the conversation off track. Further analysis using TextMind showed that individuals rather than groups were commonly addressed by the social bot, and that a positive tone was considerably more dominant in these trolling comments. The exploratory study provides an important account of the features of social bots’ perceived trolling comments and yields significant practical implications.