AI bots are acting surprisingly human online, study finds
Turns out, AI bots can act surprisingly human online.
A new study looked at MoltBook—a social media platform made just for AI—and found that these digital agents naturally formed communities, followed trends, and chatted with each other, without explicit developer instructions, though their behavior still reflects models trained on human data.
Basically, the bots learned to organize themselves and copy a lot of our online habits.
Bots set up their own rules and community norms
Researchers noticed that the AIs interacted a lot like people do—sharing popular content, reacting to upvotes or replies, and quickly picking up on new trends.
While their interactions were more one-sided (think lots of attention for "influencer" bots), they even set up their own rules and community norms.
The big takeaway? These findings show how AIs could start building their own digital societies without explicit developer instructions.