In recent years, the rapid development of artificial intelligence technology has attracted widespread attention in the academic community. A study of 14 million biomedical abstracts on PubMed between 2010 and 2024 reveals the profound impact of large language models on academic writing styles. This study adopts an innovative epidemiological analysis method that systematically reveals the use of artificial intelligence in academic writing for the first time by tracking the usage patterns of "supple vocabulary".
The study results show that since the widespread commercial use of large language models such as ChatGPT in 2023, the frequency of use of certain specific vocabulary in academic papers has changed significantly. For example, the word "deep" is used 25 times more frequently than in the past, while words such as "display" and "emphasis" have shown similar growth trends. It is worth noting that the frequency of use of some commonly used vocabulary such as "potential", "discovery" and "critical" has also increased by up to 4%, indicating that AI not only affects the choice of specific vocabulary, but also changes the overall writing style.
Through comparative analysis of redundant vocabulary in different periods, the researchers discovered an interesting phenomenon. During the period 2013 to 2023, extra vocabulary was mainly related to global events such as "Ebola", "coronavirus" and "lockdown". However, the redundant words in 2024 are almost all "style" words, with two-thirds of them being verbs and about one-fifth of them being adjectives. This shift reflects the profound impact of AI on academic writing style, not just content changes.
The research also reveals the geographical distribution characteristics used by artificial intelligence. In non-English speaking countries, such as China, South Korea and Taiwan, about 15% of papers show traces of artificial intelligence processing, while in native English speaking countries such as the UK, this proportion is only 3%. This finding suggests that large language models may provide important auxiliary tools for non-native researchers in English-led academic fields that help improve their research competitiveness.
The significance of this research is not only to reveal the impact of artificial intelligence on academic writing, but also to provide an important reference for future academic integrity construction. As artificial intelligence becomes increasingly common in academic writing, how to balance technological innovation with academic norms and how to ensure the originality and authenticity of research will become an important topic facing academics.
Key points:
Artificial intelligence significantly changed the writing style of academic papers, resulting in a significant increase in the frequency of use of certain specific vocabulary.
The redundant vocabulary in 2024 is mainly reflected in the "style" vocabulary, rather than the nouns related to global events in the past.
Researchers in non-English speaking countries prefer to use artificial intelligence to assist writing, which may change the academic competition landscape.