The aftereffects of ChatGPT are here. Human daily conversations are becoming increasingly AI-like.
After chatting with AI for more than two years, is human speech starting to sound more and more like ChatGPT?
The latest research results show that it is indeed the case.
The research team at Florida State University spent two years analyzing unscripted spoken recordings before and after the release of ChatGPT. In a dataset of 22.1 million words, they found that academic writing words like "delve" and "intricate" frequently appeared in people's daily conversations.
That is to say, the colloquial "Anyway, there are still some problems with our plan." might turn into —
"In summary, there is room for optimization in this plan."
Some mischievous netizens also gave a typical example with a lot of elements.
Without further ado, let's first take a look at how the research was conducted.
Frequent appearance of academic writing words in daily conversations
This is a research on "whether AI is quietly changing the way humans speak".
First of all, the research background is quite realistic: nowadays, whether writing a thesis or doing homework, academic words like "delve" and "intricate" are being used more and more often. Many people think it's because large models always like to use these words.
Then the question arises:
Are these changes due to people directly copying the content written by AI, or is it because AI has really influenced human language habits, making people unconsciously use these words?
To find the answer, the research team at Florida State University dug out 100,000 recordings and 22.1 million words of unscripted conversations from popular American technology and science podcasts.
They divided these contents into two groups: before the emergence of ChatGPT in 2022 (mainly from 2019 - 2021) and after (from 2023 - 2025), specifically focusing on the words that AI likes to use to see if the frequency of humans using these words in conversations has changed.
The results show that after 2022, the usage frequency of these AI - related words has indeed increased significantly. For example, words like "surpass" (exceed) and "align" (align) are used much more often than before;
However, for the ordinary synonyms used as a comparison, such as "explore" and "investigate", which are synonyms of "delve", there is no obvious change in their usage frequency. Some increase and some decrease, showing a balanced situation.
This indicates that the words used by humans in conversations are indeed gradually moving closer to AI's word - using habits and becoming more academic...
However, the research doesn't draw a definite conclusion and also mentions several key issues:
First, although these words are used more often, it's impossible to determine that it's directly caused by AI. Maybe these words were already becoming popular gradually, and AI just gave them a push;
Second, this research only focuses on technology podcasts. People who listen to these podcasts are usually in frequent contact with AI, so the results may not represent everyone;
Third, there are occasionally advertisements or pre - prepared questions in podcasts. Although the proportion of these contents is small, they may still affect the data.
Finally, the research also raises some in - depth concerns about the permeation effect: If AI's word - using habits quietly permeate into human language, what if AI has bad tendencies? Will it also affect human thoughts?
At the 8th Conference on Artificial Intelligence, Ethics, and Society to be held in October, this research will be published as part of the conference in the AIES proceedings.
Due to the rapid development of large - model capabilities, it's now increasingly difficult to distinguish whether a piece of text is written by a human or AI, which also brings trouble to the research of language itself.
To prove that they haven't used AI, netizens have also given some "little tricks": sometimes they make some small mistakes to imply that they haven't used AI.
So, have you used any AI - related words recently?
Paper address: https://arxiv.org/abs/2508.00238
Reference links:
[1]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/
[2]https://news.ycombinator.com/item?id=45045500
This article is from the WeChat public account "QbitAI". Author: Wen Le. Republished by 36Kr with authorization.