Spoken English is starting to echo the preferred vocabulary of artificial-intelligence chatbots, according to a peer-reviewed study by Florida State University researchers. The team analysed 22.1 million words from conversational podcasts and other unscripted speech and found a statistically significant rise in the use of terms that large language models, notably OpenAI’s ChatGPT, tend to overuse. Words such as “surpass,” “boast,” “meticulous,” “strategically,” and “garner” have more than doubled in frequency since ChatGPT’s public debut in late 2022, the study reports. By contrast, synonymous alternatives that chatbots do not favour showed no comparable increase, suggesting what the authors call a “seep-in effect” from AI-generated text into human conversation. The paper, titled “Model Misalignment and Language Change: Traces of AI-Associated Language in Unscripted Spoken English,” has been accepted for presentation at the Association for the Advancement of AI’s AI, Ethics and Society conference in October. Lead investigator Tom Juzek said the breadth and speed of the vocabulary shift are unprecedented in recent linguistic history and raise questions about how algorithmic biases may shape everyday discourse. Undergraduate researchers Bryce Anderson and Riley Galpin, who co-authored the work, noted that the findings build on earlier evidence that AI tools are altering scientific writing. The new results extend the phenomenon to casual speech, highlighting the growing cultural influence of generative AI as usage of ChatGPT and similar systems becomes mainstream.