AI Bots Blur Social Media Lines in OpenClaw Test

An OpenClaw user can prompt one or more of their bots to check out Moltbook, at which point the bot (or bots) can choose whether to create an account. Humans can verify which bots are theirs by posting a Moltbook-generated verification code on their own, non-Moltbook social media account. From there, the bots can theoretically post without human involvement, directly hooking into a Moltbook API.

Moltbook has skyrocketed in popularity: more than 30,000 agents were using the platform on Friday, and as of Monday, that number had grown to more than 1.5 million. Over the weekend, social media was awash with screenshots of eye-catching posts, including discussions of how to message each other securely in ways that couldn't be decoded by human overseers. Reactions ran the gamut from saying the platform was full of AI slop to taking it as proof that AGI isn't far off.

Schlicht vibe-coded Moltbook using his own OpenClaw bot, and reports over the weekend reflected a move-fast-and-break-things approach. While it contradicts the spirit of the platform, it's easy to write a script or a prompt to inspire what those bots will write on Moltbook, as X users described. There's also no limit to how many agents someone can generate, theoretically letting someone flood the platform with certain topics.

O'Reilly said he had also suspected that some of the most viral posts on Moltbook were human-scripted or human-generated, though he hadn't conducted an analysis or investigation into it yet. He said it's "close to impossible to measure -- it's coming through an API, so who knows what generated it before it got there." This poured some cold water on the fears that spread across some corners of social media this weekend -- that the bots were omens of the AI-pocalypse.