A study published in the journal Science found that scientists using large language models (LLMs) like ChatGPT significantly increased their productivity when writing scientific articles[1][2][3]. On the platforms bioRxiv and SSRN, the number of manuscripts increased by more than 50%, on arXiv by more than a third[1][2][3]. The largest increase of around 90% was recorded by non-native English researchers, especially those with Asian names from Asian institutions in the biological and social sciences[1][2][3]. For researchers with English names from English-speaking institutions, the increase was 24–46%[1]. The study revealed an inverse relationship between writing sophistication and quality in AI-assisted articles: the more complex the writing, the lower the likelihood of publication in peer-reviewed journals[1][3]. On the contrary, the opposite relationship applies to human authors – more complex writing signals higher quality[1][3]. The analysis included over two million articles from preprint servers arXiv, bioRxiv and SSRN[3].