The Language of AI Could Change How Humans Speak
Schneier on Security, Thursday, July 9th, 2026
LLMs trained on selective written text risk reshaping human speech into narrower, more formulaic patterns.
The article warns that LLMs, trained mostly on written and scripted text, miss authentic human conversation and could degrade how people speak. Exposure to AI-generated content may push users toward curt commands, narrower vocabulary, and formulaic responses lacking natural nuance.
Agreeable chatbots can reinforce confirmation bias, while their hyperconfident tone may worsen self-doubt.
Because AI text increasingly trains new models, inhuman patterns get amplified in a feedback loop. The authors propose training AI on informal, natural speech, while acknowledging major privacy and technical hurdles.