As artificial intelligence becomes embedded in everyday work and communication, researchers and professionals are noticing an unexpected reversal in the human-machine relationship: instead of AI merely learning to imitate people, people appear to be adopting the vocabulary and rhetorical habits of AI. Workers report catching themselves using chatbot-associated constructions such as contrast framing, excessive politeness, formalized requests and tightly structured “prompt speak” in meetings, texts and ordinary conversation. Research reinforces the anecdotes, finding substantial increases in words disproportionately favored by large language models—including “delve,” “showcase,” “boast,” “intricacies” and “meticulous”—in human speech following ChatGPT’s release. Other research has found that people can absorb a chatbot’s lexical preferences after relatively brief interactions. The emerging phenomenon illustrates a potentially important cultural feedback loop: AI systems were trained on human language, but as people consume enormous quantities of AI-generated material, the linguistic patterns produced by those machines may increasingly flow back into human vocabulary, workplace communication and eventually the broader culture.
Key Takeaways
- Researchers have found measurable increases in human use of vocabulary strongly associated with large language models, providing evidence that exposure to AI-generated language may be influencing how people actually speak.
- The influence extends beyond individual words to communication habits, with workers reporting “prompt speak,” contrast framing, hyper-polite requests, formulaic organization and other patterns commonly encountered when interacting with ChatGPT and Claude.
- The trend creates a potentially self-reinforcing linguistic cycle: AI models learn from human language, humans adopt language generated by AI, and future AI systems may subsequently be trained on a culture increasingly influenced by earlier AI output.
In-Depth
Artificial intelligence was designed to communicate more like human beings. An intriguing reversal is now emerging: human beings are beginning to communicate more like artificial intelligence.
The phenomenon goes beyond people deliberately copying chatbot responses. Workers increasingly report catching themselves using AI-associated sentence structures and vocabulary in meetings, messages and ordinary conversations. Researchers have uncovered supporting evidence in large collections of spoken language, finding noticeable increases in words disproportionately favored by ChatGPT after its public introduction.
This should not automatically be treated as linguistic decline. Language has always evolved through imitation, technology and cultural exposure. Radio, television, advertising, corporate culture and social media have all influenced how Americans speak. AI could similarly introduce useful precision, structure and efficiency into communication.
But AI presents a fundamentally different scale. Millions of people repeatedly interact with systems whose linguistic output is generated from statistical patterns and reinforced through standardized model behavior. If those patterns migrate into everyday speech, linguistic diversity could gradually narrow as people unconsciously adopt similar vocabulary, sentence construction and rhetorical conventions.
There is also a deeper cultural question. Human language reflects personality, geography, profession, family, education and lived experience. Those differences give communication much of its character. Efficiency becomes a questionable bargain if everyone begins sounding as though their sentences were produced from the same template.
The challenge, therefore, is not rejecting AI assistance. It is maintaining human authorship over human communication. AI can sharpen language, but people should remain conscious of preserving individual voices rather than allowing technological convenience to quietly standardize them.
Sources
- https://www.wsj.com/tech/ai/claude-chatgpt-ai-speak-chatbot-cb804933
- https://arxiv.org/abs/2409.01754
- https://news.fsu.edu/news/education-society/2025/08/26/on-screen-and-now-irl-fsu-researchers-find-evidence-suggesting-chatgpt-influences-how-we-speak/
- https://www.scienceopen.com/document?vid=e4f9d89e-55ce-4246-97e7-ce78505b6fc1
- https://pubmed.ncbi.nlm.nih.gov/42193597/