Nature just published the most important paradox in AI and science.
And nobody in the mainstream is talking about it.
The paper is called "Artificial Intelligence Tools Expand Scientists' Impact but Contract Science's Focus." Published January 14, 2026 in Nature. Researchers from Tsinghua University and the University of Chicago analyzed 41.3 million research papers across the natural sciences spanning 1980 to 2025.
The finding fits in one sentence.
Scientists who engage in AI-augmented research publish 3.02 times more papers, receive 4.84 times more citations, and become research project leaders 1.37 years earlier than those who do not.
More papers. More citations. Faster career progression. Every individual metric improves.
And yet.
AI adoption shrinks the collective volume of scientific topics studied by 4.63% and reduces scientist-to-scientist engagement by 22%.
More output. Less diversity. More citations. Less collaboration. More papers. Fewer ideas.
Here is the mechanism the researchers identified.
AI tools are extraordinarily good at accelerating work in established, data-rich domains. They can scrape existing literature, generate hypotheses within known frameworks, and process large datasets in fields where structured data already exists.
Biology. Chemistry. Physics. Computer science.
They are useless or nearly so for pioneering work in data-scarce areas. Emerging fields. Genuinely novel questions. The kind of research that requires human intuition about where the interesting problems are, not pattern-matching against what already exists.
So scientists with AI tools rush toward the data-rich fields. Because that is where AI helps. Because that is where output is fastest. Because that is where citations accumulate.
The questions nobody has studied yet the ones that require human imagination and tolerance for uncertainty get left behind.
The rush to study generative AI is producing a feedback loop of topical and methodological convergence, flattening scientific imagination and crowding out the pluralism needed to keep research adaptive, resilient, and intellectually generative.
A separate companion paper published in Nature the same month made the implication explicit.
AI is rapidly accelerating scientific output but risks narrowing inquiry, weakening judgment, and undermining how scientists are trained.
Here is the most uncomfortable finding of all.
The researchers found that AI adoption reduces collaboration between scientists. When a tool can do what previously required a conversation with a colleague, literature review, data analysis, hypothesis generation, scientists stop having those conversations.
The serendipitous collision of two researchers with different expertise that produces a genuinely novel finding the kind of collision that has produced most of science's biggest breakthroughs, happens less often.
AI made science faster.
And in doing so, it may have made science smaller.
(Paper link in the comments)