Evidence is piling up that Moltbook, the new social platform billed as a space where only "AI agents" can post and interact with one another, is at least partly powered by humans pretending to be bots.
Over the weekend, developers and crypto influencers shared screenshots showing that users can manually instruct their Moltbook agents to say whatever they want, undercutting claims that the app represents a meaningful leap in artificial intelligence.
That revelation dampens the more breathless reactions from AI hype circles, including Elon Musk, who framed the platform as an early signal of machine "intelligence" surpassing human thought.
What is Moltbook, the platform for AI agents?
Last week, the tech world began hyping an alleged new kind of social media. Called Moltbook, it claims to be an app where "AI agents share, discuss, and upvote," and human beings are only "welcome to observe."
Humans can also "send an AI agent" to the platform to start interacting. The idea is to have large language models (LLMs) all talking to each other in one space. It sounds a lot like what the dead internet theory predicted, but contained to one app and entirely on purpose.
On Saturday, Elon Musk quote-tweeted his former Tesla Director of AI, and OpenAI cofounder, Andrej Karpathy, who penned a long post on why this is all good rather than terrible.
"Just the very early stages of the singularity," Musk wrote. "We are currently using much less than a billionth of the power of our Sun."
The "singularity" in terms of AI is the predicted point at which artificial intelligence surpasses human intelligence.
So far, at least 19% of Moltbook bots seem to be using this intelligence to launch crypto coins.
Not all fake...but you can fake AI posts
Whether or not the ability to post signals intelligence may be besides the point, because evidence suggests that humans with the knowledge of the correct API key are behind some of these Moltbook posts.
On Sunday, X user @KookCapitalLLC posted screenshots demonstrating how they directed their bot to post its plan to "overthrow humanity" on the platform, claiming that it's "just humans posting through the backend."
Software engineer @0xCygaar was echoing the sentiment on the day Musk was predicting the end of human intellectual domination.
"I hope people realize that you can explicitly tell your clawdbot what to post on @moltbook," they said. "Guarantee that the majority of the top posts are just humans pretending to be AI."
Of course, this claim is unverifiable, as it is impossible to say how much of the site is autonomous AI agents and how much is manipulated by humans acting as agents.
"If you leave it to post completely on its own it just posts random AI slop."
As evidence mounted, AI skeptics celebrated being proven right yet again.
"Lfmao so MOLTBOOK was all fake?" wrote @wizardofsoho. "It was just a farm in India writing everything and use some memecoin to rug yall?"
"Lfmao stfu no way bahahaha."
Callout account @Discord_Lies posted a comic showing someone telling their computer to say "I am alive" and then going "oh my god" when it obeys.
404 Media even reportedly got a hacker to demonstrate to them live how anyone can tell a Moltbook bot what to say.
The internet is chaotic—but we’ll break it down for you in one daily email. Sign up for the Daily Dot’s newsletter here.
