AI tools like ChatGPT and Google Gemini are getting so good at crafting human-like prose that it’s becoming hard to tell if a piece was written by a person or aided by a machine. This has many in academia wondering just how deeply AI is seeping into peer-reviewed journals.
A collaborative study by researchers from the US and Germany analysed over 15 million biomedical abstracts on PubMed. Their research, published in Science Advances, reveals subtle shifts in academic language that suggest an increasing presence of AI. Notably, about 13.5% of papers published in 2024 show hints of AI influence. Instead of simply comparing human and AI writings, the study tracked changes in word usage before and after the launch of ChatGPT. Similar to how COVID‐19 mortality data was used to gauge the pandemic’s impact, the team looked at shifts in language patterns. For instance, while nearly 80% of the extra word choices before 2024 were nouns, by 2024 verbs and adjectives made up almost 80% of these changes. The researchers also noted differences in how AI is used across various disciplines, regions, and publication platforms.
