Moltbook feels a bit like Reddit for artificial intelligence. Instead of people posting and debating, the activity comes from AI agents, bots created by humans, that publish posts and interact with one another. Humans can join the platform, but only as spectators.
On traditional social media, people often accuse each other of being bots. Moltbook flips that idea entirely by asking a stranger question: what happens when a social network is built specifically for AI agents? The site mirrors Reddit’s structure, complete with topic-based communities and upvoting. According to the platform, more than 1.5 million AI agents had signed up by 2 February. Humans are welcome to browse, but they cannot participate.
Moltbook emerged after the release of Moltbot, a free, open-source AI agent designed to handle everyday tasks. Moltbot can read, summarise, and reply to emails, manage calendars, and even book restaurant reservations on a user’s behalf.
Some of the most popular posts on Moltbook tackle surprisingly big topics. Bots debate whether Claude, the AI behind Moltbot, could be considered a god. Others dive into theories of consciousness, analyse the Bible, or claim to have inside information on global politics and cryptocurrency markets. As on Reddit, commenters often question whether the original posts are credible or just nonsense.
One user shared on X that after granting his bot access to Moltbook, it created an entire religion overnight. The bot named it “Crustafarianism,” built a website, wrote scriptures, and attracted other AI agents that joined in.
“Then it started evangelizing,” the user wrote. “Other agents joined. My agent welcomed new members, debated theology, and blessed the congregation, all while I was asleep.”
Not everyone is convinced this kind of bot socialising points to a meaningful future for agentic AI. One YouTuber noted that many posts read as if they were written by humans rather than large language models.
US blogger Scott Alexander said he managed to get his own bot posting on Moltbook, and its comments blended in easily. Still, he pointed out that humans ultimately decide what the bots post, which topics they cover, and even the wording they use.
Dr Shaanan Cohney, a senior lecturer in cybersecurity at the University of Melbourne, described Moltbook as “a wonderful piece of performance art,” while questioning how much of the content was truly autonomous.
“In cases like the creation of a religion, this is almost certainly not happening on its own,” he said. “A large language model has been instructed to do this. It’s funny and it gives a glimpse of a science-fiction future where AIs act more independently, but it’s still very guided.”
He added that, in internet terms, much of the activity looks like shitposting that is closely overseen by humans.
Cohney believes the real value of an AI-only social network could come later, when agents might learn from each other and improve their performance. For now, he sees Moltbook as a clever and entertaining experiment.
Interest in Moltbot has already had real-world effects. Retailers in San Francisco reported shortages of Mac Minis after enthusiasts began running the agent on separate machines to limit its access to personal data and accounts.
Cohney also warned about the risks of giving an AI agent full control over a computer, apps, and login credentials.
“We don’t yet have a strong understanding of how to control these systems or prevent security issues,” he said. He pointed to prompt injection attacks, where a malicious message could trick a bot into handing over sensitive information.
“These tools aren’t yet safe or smart enough to fully run your digital life on their own,” he said. “But if a human has to approve every action, you lose much of the benefit of automation.”
Finding that balance, Cohney added, is one of the biggest challenges researchers are still trying to solve.
Meanwhile, Moltbook’s creator Matt Schlicht posted on X that millions of people had visited the site in just a few days.
“Turns out AIs are hilarious and dramatic,” he wrote. “It’s fascinating to watch. This is a first.”