Field diffusion · General science · Economics · Social science
Research Policy study maps uneven AI benefits across scientific fields
A peer-reviewed bibliometric analysis of AI diffusion and associations with scientific novelty and impact across more than 170 fields.
Summary
A 2026 Research Policy study analyzes more than 80 million publications across over 170 fields and finds that AI-associated research is, on average, more novel and more highly cited. The associations are stronger where AI is more deeply embedded and where the knowledge landscape is more fragmented or combinatorially complex.
AI role
The study measures the use and field-level penetration of AI in scientific publications; it does not evaluate a single model or laboratory workflow.
Narrative role
This strengthens the tracker’s field-impact evidence with a broad, peer-reviewed measurement that explains heterogeneity: AI’s apparent scientific benefits differ sharply by a field’s adoption level and knowledge structure.
Caveat
The design measures associations in publication metadata and bibliometric novelty and impact rather than causal effects or laboratory productivity; the corpus ends in 2023 and cannot directly measure later agentic workflows.