Field diffusion · General science · Social science · Computer science

Study quantifies AI use and citation-associated benefits across science

A peer-reviewed measurement framework for direct AI use and citation-associated benefits across 19 scientific disciplines and 292 fields.

Summary

Gao and Wang developed a text-based framework to measure direct AI use in scientific publications and its association with citation impact. Across a large historical corpus, they find rapidly increasing use since 2015 and widespread but uneven citation-associated benefits, alongside gaps between AI training and research use and demographic disparities across disciplines.

AI role

AI methods are identified from publication text and their use is compared with subsequent citation impact within and outside disciplines.

Narrative role

This fills a historical field-impact gap with an early peer-reviewed, science-wide measurement of both AI diffusion and heterogeneous research benefits, providing a baseline for the tracker's later OpenAlex, AlphaFold, and post-LLM studies.

Caveat

AI use is inferred from publication text and benefits are measured through citations, so the observational associations do not establish causal productivity or discovery effects; the corpus ends before widespread generative-AI use.