Generative AI use in radiology manuscript writing has more than doubled
Generative artificial intelligence utilization in radiology manuscripts is on the rise, and publications need to adapt accordingly, the editor-in-chief of a prominent radiology journal cautions.
A new paper in the American Journal of Roentgenology, written by Andrew B. Rosenkrantz, MD, details the emergence of generative AI in manuscript writing. In it, Rosenkrantz shares insight into how the submissions to AJR have changed since March of 2024 when the journal added a question in its online manuscript submission interface asking authors whether generative artificial AI was used in the process. Answering the question is a requirement to complete the submission process, and a free-text follow-up question is available for authors to disclose additional information.
Since then, the journal has received nearly 4,000 manuscripts. Of those, around 3% disclosed the use of the technology, but these figures increased significantly between 2024 and 2026. In 2024, around 1.1% of manuscripts included generative AI content in some form. By 2025, this figure more than doubled, and half way through this year, the figure had increased significantly yet again, with 7.5% of all manuscripts utilizing generative AI in some way.
The technology was used the most for full-length research and reviews; of those submissions, around 4.4% that used AI were accepted, compared to 5% of those that did not disclose AI use. Artificial intelligence was most often used for editing to ensure readability and conciseness, check errors and adhere to journal-specific formatting requirements. Researchers most often achieved this with the help of ChatGPT (59.7%), Claude (8.4%) or Gemini (6.7%).
Rosenkrantz noted that the figures observed in AJR’s data are likely an underestimation of generative AI use in manuscripts. He suggested that many authors might still be hesitant to admit to utilizing AI due to perceived stigma and the potential negative impact it could have on the peer review process.
“The use of generative AI in scholarly writing carries well-recognized risks including loss of originality, perpetuation of biases, and creation of inaccurate or misleading information,” Rosenkrantz wrote July 8. “However, it also has benefits, such as helping to improve manuscript clarity, mitigate disparities relating to language fluency, and ultimately ‘democratize both the creation and consumption of scientific knowledge.’ Moreover, analyses suggest substantial increases in productivity from the use of generative AI for writing (40% time savings) and markedly increased scientific output by researchers who adopt generative AI compared with those who do not.”
He cautioned, however, that its use does not come without risks for authors and publications alike. He warned that disclosure may soon not be enough to ensure the integrity of manuscripts, and that any authors who choose to utilize AI should do so judiciously and with integrity.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.