AI Chatbots Overuse 'Not X but Y' Phrasing, Tripling Its Rate in Machine-Written Text
Updated
Updated · The Atlantic · Jul 12
AI Chatbots Overuse 'Not X but Y' Phrasing, Tripling Its Rate in Machine-Written Text
1 articles · Updated · The Atlantic · Jul 12
Summary
Pangram estimates “not just X but Y” constructions appear three times as often in AI writing as in human text, making the pattern one of the clearest persistent chatbot tells.
Barron’s reported the phrasing more than quadrupled in corporate communications from 2023 to 2025, suggesting the habit is spreading beyond chat interfaces into business writing.
OpenAI says ChatGPT leans on the device too often and is trying to widen its stylistic range, while users already swap prompts and editing tricks to strip the pattern from outputs.
Researchers say the overuse likely stems from a mix of human training data, reinforcement learning that rewards “nuanced” answers, and token-by-token prediction that makes the structure a safe next-word path.
That feedback loop may be hard to break because newer models train on AI-generated text, raising broader worries about model collapse and AI writing tics bleeding into human speech.
As humans adopt AI's robotic language, are we trading creativity for machine-like efficiency?
In a world of perfect AI text, how can anyone prove their own words are authentically human?
Formulaic AI Language Quadruples in Corporate Reports: Detection, Risks, and the Future of Human Communication
Overview
As AI models become increasingly sophisticated, a new phenomenon called the 'AI tell' has emerged—subtle linguistic patterns or stylistic choices that can inadvertently reveal machine authorship. Identifying these tells is now a critical focus for researchers and content creators, especially as AI-generated text spreads across many domains. One prominent example is the 'Not X, but Y' phrasing, which has appeared with growing frequency and creates a familiar ring in corporate communications. Notably, the use of this phrasing in such settings more than quadrupled between 2023 and 2025, highlighting the measurable impact of AI on modern writing.