Scientists Warn of a Creepy New AI Side Effect on Humans
New research suggests the more people interact with AI chatbots, the more they start to sound like one themselves.
The list of AI worries was already long — job losses, security threats, "AI psychosis," eroding skills, deeper isolation, an endless stream of low-quality "slop" content, even fears about AI-assisted bioweapons. Now there's one more to add: talking to chatbots may be making us talk like chatbots.
How AI conversations change us
Think about the last time you dealt with an AI customer service bot. If it left you frustrated, it probably wasn't because the AI was unhelpful or unfriendly — those bots are almost always polite. It was more likely because the AI couldn't grasp some very human nuance in your request.
So what did you do? According to new theoretical research from an international team of scientists, you probably simplified. You dropped the slang, the metaphors, the personal detail. You made your request explicit, direct, and stripped of anything uniquely "you."
In other words, you started sounding a little more like a machine.
A two-way training process
The study, published in AI & Society, argues this isn't a one-off — it's a feedback loop. AI systems are well known to be learning from human language and becoming more conversational. But the researchers say the reverse is happening too: every time a person gets what they need by talking like a robot, that behavior gets reinforced, nudging them to do it again next time.
Linnaeus University professor and study co-author Selcen Ozturkcan describes it as a mirroring effect — people naturally tend to match the communication style of whoever (or whatever) they're talking to, and repeated exchanges with AI can gradually shift how people speak more generally.
As AI grows more human and humans adapt to AI, the researchers say the two are converging toward what they call "robotoid humanness" — a shared, flattened communication style that isn't quite either.
Should we be worried?
The paper is theoretical — the team hasn't run experiments proving this is already happening. But it builds on earlier work suggesting AI interaction may be reshaping how people think and express themselves more broadly.
The underlying logic isn't new: people respond to incentives. Writers dropped certain phrasings once readers started flagging them as "AI-sounding." Employees lean harder into AI tools when using them earns praise. By the same reasoning, if speaking more like a bot is what gets results, people may keep speaking more like a bot.
It's not on the same scale as more dramatic AI risks — but it's a quietly unsettling thought: the more we optimize our language for machines, the less room we may leave for the messiness that makes us human.