Our new study on the Rise of Generative Artificial Intelligence in Science shows that Generative Artificial Intelligence (GenAI) is playing an increasingly significant role in research, with growing use across disciplines.
Using scientific publication data for 2017–2023 from the OpenAlex bibliometric database, we applied a two-stage classifier built on GPT-4 and SciBERT to distinguish between studies that apply AI and those that only discuss it. This enabled us to track the growth, geography, use modes, and collaboration patterns of GenAI adoption compared to other AI technologies.
Key findings include:
- GenAI-related publications have risen sharply, with applications extending beyond computer science into many fields.
- U.S. researchers produced nearly 40% of publications applying GenAI, followed by China with almost one-third; several smaller advanced economies also show high uptake.
- While more common in social sciences, GenAI discussion papers appear across the research spectrum.
- GenAI application papers are linked to larger teams and greater international collaboration. Application-oriented research with GenAI, possibly due to its novelty, may require additional technical and domain-specific expertise. Even GenAI discussion papers show a positive link to cross-border collaboration
Our classifier provides a finer-grained view of whether researchers are using GenAI tools or reflecting on their implications. The results underline GenAI’s growing role in shaping science, with implications for policy and international cooperation. Limitations include definitional choices, data coverage, and the difficulty of capturing all uses of GenAI from publication metadata.
For details, read the full paper (open access): Ding, L., Lawson, C. & Shapira, P. Rise of Generative Artificial Intelligence in Science. Scientometrics (2025). https://doi.org/10.1007/s11192-025-05413-z
Philip Shapira 