AI
ChatGPT is reshaping how people speak, research shows
A new study finds that words favored by large language models are seeping into everyday human conversation, suggesting AI chatbots are subtly shaping cultural language patterns.
Language models like ChatGPT may be quietly influencing how humans talk. Researchers have documented measurable shifts in everyday speech toward words that these AI systems preferentially generate, suggesting a feedback loop between artificial intelligence and human communication.
The linguistic shift#
Analysis of nearly 737,000 hours of unscripted podcast conversation across over 820,000 episodes revealed that words commonly favored by ChatGPT—including delve, showcase, boast, intricacies, and meticulous—began appearing more frequently in human speech after the chatbot's release. Using a synthetic-control methodology, the researchers established a causal link between ChatGPT's debut and this sudden uptick in vocabulary adoption, suggesting the influence extends beyond chance or coincidence.
How people pick up AI language#
A controlled experiment with nearly 500 participants demonstrated the mechanism behind this shift. When people briefly interacted with a chatbot, they absorbed and began using its characteristic word choices. Notably, these borrowed words persisted even after participants completed distractor tasks, and they continued appearing in subsequent word choices—evidence that the vocabulary becomes anchored in people's active speech.
Broader implications#
This finding highlights a novel dimension of cultural evolution: machines trained on human language are now feeding their distinctive patterns back into human discourse at scale. The concentration of this influence in the hands of a few major AI companies raises questions about linguistic homogenization and the potential for latent cultural influence through language choices embedded in widely-used systems.
