A new analysis suggests AI tools are shaping most biomedical manuscripts. A preprint posted to arXiv estimates that 90% of PubMed Central papers published in December 2025 showed linguistic signatures of large language models, with usage at 77% for all of 2025 and 52% for 2024—well above earlier abstract-only estimates. The authors attribute the jump to a more sensitive detection method and report heavier AI use in narrative sections, such as abstracts and discussions, than in methods and results, raising concerns about bias and hallucinated findings. Outside experts say the trend tracks with widespread adoption but caution the sample may not reflect all of science. Either way, AI-assisted writing appears entrenched in the research workflow.
Related articles:
— Science sleuths flag hundreds of papers that use AI without disclosing it
— Three ways ChatGPT helps me in my academic writing
— Signs of AI-generated text found in 14% of biomedical abstracts last year
— Is it OK for AI to write science papers? Nature survey shows researchers are split
— Can Anthropic’s invisible watermarks curb ‘AI slop’? Researchers remain sceptical
