Simon Willison has introduced a new investigative tool aimed at identifying automated reply bots on the Bluesky social network. Built using a “vibe coding” approach with the Opus 5.5 model, the tool allows users to examine specific profiles for indicators of automated activity. This development comes as Bluesky begins to experience the same types of automated swarms that have long affected other major social platforms like Twitter.
Addressing Automation on Decentralized Platforms
The emergence of reply bots on Bluesky represents a shift for the platform as it gains more users and visibility. Willison reported that while automated accounts are becoming more frequent, Bluesky’s technical architecture offers a distinct advantage for those attempting to mitigate them. Unlike Twitter, which moved to a restricted and paid model for its data access, Bluesky maintains a freely available and functional API. This open access does not prevent bots from operating, but it significantly lowers the barrier for developers to build counter-tools and investigate suspicious accounts. The tool leverages this accessibility to provide evidence of likely bot behavior, a task that has become increasingly difficult on platforms with closed ecosystems.
Unknown Implementation Details
Several aspects of the tool’s functionality and deployment remain unaddressed. The announcement did not specify the specific metrics or behavioral patterns the tool uses to classify an account as a “likely reply bot.” It is also unknown whether the tool is available as a public web application or if the source code has been released for independent hosting. Furthermore, the report did not provide details on the accuracy rates of the Opus 5.5-generated code or how it distinguishes between sophisticated human users and high-frequency automated scripts. As the platform evolves, it remains to be seen if Bluesky will implement native solutions or continue to rely on community-developed tools to manage its bot population.
This bot checker serves as a practical application of large language models in social media moderation. It highlights how open data access can empower users to address platform-wide issues independently.
Lead image: Bibek ghosh / Pexels
