Can AI make scientists faster without making science narrower? The newest evidence suggests: yes, and that is the problem. A Nature study published on January 14, 2026, examined 41.3 million research papers and found that scientists using AI published 3.02 times more papers, received 4.84 times more citations, and became project leaders 1.37 years earlier than scientists who did not use AI. But the same study also found a cost: the total range of topics being studied shrank by 4.63%, and engagement between scientists fell by 22%. In other words, AI seems to help individuals move faster, while pushing science as a whole toward safer, data-rich areas instead of unexplored ones. (nature.com)
This concern is not only about speed. It is also about who benefits. A 2024 paper in Nature Human Behaviour reported that AI use and its benefits have spread widely across many fields, especially since 2015. However, it also warned of a gap between AI education and real research needs, and found that fields with higher proportions of women or Black scientists appeared to gain fewer benefits from AI. That means AI could deepen old inequalities instead of solving them. Another sign of concentration comes from life-science research: a 2024 Nature Communications study found that the United States and China produced almost half of the world’s AI life-science research from 2000 to 2022, while Latin America and Africa lagged far behind. (nature.com)
Still, the story is not simply negative. AI language tools can help researchers who are not fluent in English, which has long been a major barrier in global science. Yet even here, the picture is mixed. A 2025 Scientific Reports study found that AI-generated content in Arabic and Chinese was less complete, actionable, and creative than in English, with especially weak results in technical R&D tasks. So AI may open one door while quietly closing another. The real question is no longer “Will AI change science?” It already has. The better question is whether universities, journals, and governments can guide AI so that it expands science for more people, instead of narrowing it to the loudest languages, richest datasets, and strongest research powers. (nature.com)










