AI Use in Federal Grant Proposals Linked to Less Distinctive Research Ideas
A new study published in PNAS reveals that scientists utilizing generative artificial intelligence (AI) to assist in writing federal grant applications tend to propose research topics that are less distinctive and more closely resemble previously funded projects. The study, conducted by Yifan Qian and colleagues at Northwestern University, analyzed grant proposals and awards from 2021 to 2025, including confidential university submissions to the National Science Foundation (NSF) and the National Institutes of Health (NIH), as well as public federal awards. The researchers found that higher AI involvement in a scientist's work was consistently associated with lower semantic distinctiveness, positioning projects closer to what the agencies had funded in the prior year. At NIH, proposals with greater AI traces were more likely to secure funding and resulted in modestly more publications in the initial one to two years, primarily routine studies rather than highly cited breakthroughs. AI involvement in grant a...