ChatGPT is changing the way we communicate — here's how you can avoid speaking like AI in public (www.techradar.com)

ChatGPT and other large language models, originally trained on vast amounts of human text, are no longer just mimicking our voice — they’re beginning to reshape it. A Max Planck Institute study identified a “measurable and abrupt increase” in words and phrasings that ChatGPT preferentially generates since its launch, evidence of a “closed cultural feedback loop” in which AI-trained patterns propagate back into everyday language. Socio-tech experts call this an “echo effect”: frequent exposure to AI-produced vocabulary (think “pivotal,” “realm,” or stocky phrases like “it is pivotal to note”) makes those terms feel natural, so people adopt them in writing and speech, from blog posts to meetings and LinkedIn updates. The consequence is a gradual homogenization of style. Because LLM outputs tend to be smooth, neutral and professional, they can iron out regional quirks and individual originality, potentially eroding linguistic and cultural diversity. Technically, the loop arises when models trained on human corpora generate widely consumed text that then influences human-produced data — which future models may be trained on. The antidote is deliberate: preserve variety by drafting in your own voice, use AI mainly for editing, and consciously swap buzzwords for simpler or idiosyncratic phrasing to keep human expression diverse and authentic.