AI is imperceptibly changing people: WSJ reveals an unusual effect

The Wall Street Journal reports that regular use of AI chatbots, such as ChatGPT and Claude, is altering people's everyday speech. Users are adopting expressions, sentence structures, and questioning styles characteristic of neural networks, which manifests in work meetings and ordinary conversations. Examples include spontaneous use of contrasts and overly approving assessments, as well as the emergence of "caveman language"—extremely short and straightforward phrases.

The Wall Street Journal reports on an unusual effect that artificial intelligence is having on people's everyday speech. Regular use of chatbots, such as ChatGPT, Claude, and Gemini, leads users to adopt expressions, sentence structures, and questioning styles characteristic of neural networks. These changes appear not only in correspondence but also during work meetings and ordinary conversations. An example is given of Sarah Wickermans, co-founder of a venture capital firm in Stockholm, who spontaneously said during a meeting: "This is not a morning meeting, it's a reboot." Such constructions, built on contrasts, are often found in responses from language models. It is also noted that short enthusiastic assessments, such as "brilliant" or "you're so smart," reminiscent of the overly approving style of AI assistants, are increasingly appearing in work conversations. These speech features have been unofficially dubbed "LLM-isms." The opposite trend is also spreading—"caveman language," characterized by extremely short and straightforward phrases that more closely resemble commands to a computer program. In one San Francisco startup, employees devised a way to combat such habits: they note instances when colleagues talk to each other as if to AI, and the loser pays for dinner. The cause of the spread is linked to the characteristics of language models, which use clearly structured responses, enumerations, and emphatically polite formulations. At the same time, WSJ interlocutors note that many of these expressions existed long before the advent of ChatGPT in business literature and corporate correspondence.