FSU researchers find evidence of ChatGPT buzzwords turning up in everyday speech

🔗https://shre.ink/ChatGPT-buzzwords-in-everyday-speech

by McKenzie Harris

The article uncovers a fascinating shift in how we speak. Researchers from Florida State University’s Departments of Modern Languages and Linguistics, Computer Science, and Mathematics analyzed whether certain AI‑style buzzwords are making their way into unscripted, spoken conversation.

Their study, titled “Model Misalignment and Language Change: Traces of AI‑Associated Language in Unscripted Spoken English,” marks the first peer‑reviewed investigation into whether widespread use of chat‑based LLMs like ChatGPT is influencing the human conversational language system. Accepted at the Conference on AI, Ethics, and Society, the findings underscore a convergence between human word choice and the patterns favored by AI models, especially after ChatGPT’s.

The team analyzed over 22 million words from unscripted, spontaneous spoken sources—namely science and technology podcasts—to compare language before and after ChatGPT’s emergence. They discovered a striking uptick in usage of AI‑associated words like “delve,” “intricate,” “surpass,” “boast,” “meticulous,” “strategically,” and “garner,” while simpler synonyms saw no similar increase.

For example, speakers increasingly used “underscore”—a word frequently appearing in ChatGPT outputs—rather than “accentuate,” its more common synonym. Nearly three‑quarters of the targeted AI buzzwords surged in usage, some even more than doubling, in unscripted human speech.

Tom Juzek, assistant professor and principal investigator, noted the breadth and rapidity of this shift is remarkable—remarkable especially because many of these words belong more to formal or academic language than everyday conversation. Bryce Anderson, lead undergraduate researcher, emphasized that AI’s influence is not confined to tools alone—it is shaping how we express ideas through language, a societal implication with wide reach.

Riley Galpin, another team member, pointed to ethical concerns: as LLMs nudge our linguistic habits, biases or misalignments in models may subtly permeate human discourse. This builds on FSU’s earlier findings about AI influence on scientific writing, now confirmed in spoken language as well.

Looking ahead, researchers remain cautious about causation. Is AI merely accelerating existing trends or molding new speech patterns? Future work must disentangle these possibilities. Still, the phenomenon—what they term a “seep‑in effect”—signals that AI may indeed be quietly reshaping human communication