Beta The Briev beta is out. Free on iPhone via TestFlight — install it in under a minute.

Join the beta ↗
Briev
Live
Technology

Study finds AI assistance in nearly nine-tenths of recent biomedical papers

A new analysis suggests that about 90% of biomedical articles published in December 2025 show evidence of large-language-model assistance.

An analysis of English-language biomedical papers archived in a leading repository indicates that AI-driven language models assisted in writing roughly nine out of ten articles released in December 2025. Extending the scope to the whole of 2025, the study estimates a 77% AI-assistance rate, up from 52% for papers published in 2024, suggesting a steep upward trend. The authors employed a refined detection technique that scans for word frequencies typical of large language models, providing direct usage estimates rather than mere lower limits.

This approach raised the estimated AI presence in 2024 abstracts from a previous 13.5% to 31%. The findings align with a 2025 survey where 71% of researchers reported using AI for writing support, implying that actual usage may exceed self-reported levels. Commentators note that while the high rates are plausible given widespread AI adoption, the sample may not represent all scientific fields and further study is needed. Concerns were raised about AI-generated content in methods and results sections, where hallucinated data could undermine research integrity.

Why it matters

The prevalence of AI in scientific writing could reshape research quality and integrity across biomedicine.

In this story

AI assistancelarge language modelsbiomedical papersPubMed Centralresearch integritysurveytext analysisAI prevalence
Get the beta ↗