By the end of last year, nine in ten biomedical papers carried the fingerprints of a language model
A preprint posted to arXiv on 11 August by Lena Holzwarth, Rita González-Márquez and Dmitry Kobak estimates that 89 percent of the open-access biomedical papers in PubMed Central published by the end of December 2025 show an excess of vocabulary associated with large language models. The method deliberately avoids AI-text detectors, which are unreliable. Instead it tracks how the frequency of particular words shifted across the whole corpus after chatbots became widely available, and measures the surplus. That is more sensitive than earlier attempts, which is why this figure sits above previous estimates. The breakdown by section is the informative part: discussion sections came in at 68 percent, abstracts 67, introductions 59, results 46 and methods 32 — heaviest where a paper argues, lightest where it reports what was actually done. Nature covered the preprint on 21 August. It has not yet been peer reviewed, and the measure picks up editing and polishing as well as drafting; it cannot tell you which of the two happened.
Related stories
Roughly 90 percent of executives say AI has not raised productivity. The cuts are happening anyway.2026-08-24An older Claude broke Anthropic's own content rule in 10 tries out of 10 — and it is still sold through Azure and Bedrock
2026-08-23When a small model hands a conversation to a big one, the big one re-reads everything. Nvidia says a straight line can do the translation instead.
2026-08-23A 27-billion-parameter model beat Opus 4.8 and GPT-5.5 at reproducing published papers
2026-08-23Give a model the reference list of an unpublished paper and ask what the paper says. The best ones get it 15 percent of the time.
2026-08-22