Willison Builds Bluesky Bot Detector Using Claude Opus 5.5

On September 27, 2026, developer Simon Willison released a new web tool designed to analyze Bluesky profiles and identify automated reply bots. Willison created the application using a development method he described as vibe coding, powered by the Claude Opus 5.5 large language model. The tool arrives as Bluesky experiences an influx of automated spam accounts, a phenomenon that has long plagued older social networks like Twitter but is now manifesting on the newer decentralized platform.

Unlike Twitter, which has restricted its developer platform, Bluesky maintains a freely available and open API. While the lack of an open API on other platforms does not stop bot creators, it severely hinders developers trying to investigate them. Bluesky's open API allowed Willison to quickly build a tool that queries the network's data directly to evaluate specific behavioral signals and flag suspicious accounts.

The detector looks for several key indicators of automation. It flags profiles that post replies within seconds of other posts from the same account, as well as accounts that never publish original content, images, or links, but instead focus entirely on replying to high-follower users. Additionally, the tool scans for the presence of question marks in replies, a common tactic used by engagement-baiting bots to trick human users into responding.

For software practitioners and community managers, this project highlights how modern AI models like Claude Opus 5.5 can rapidly generate functional utility applications to combat platform abuse. By leveraging open APIs alongside advanced LLMs, developers can deploy targeted moderation tools far faster than traditional development cycles allowed, offering a blueprint for rapid-response community defense.