This paper examines how artificial intelligence is associated with shifts in scientific production following the sudden release of AlphaFold2, a deep-learning system that predicts protein structures with near-experimental accuracy. Its public release in 2021 created an openly accessible and field-wide change in the cost and feasibility of working on protein-related research problems. I construct a new protein-level dataset linking proteins to their experimental structural information, AlphaFold2 coverage, and the full corpus of their associated scientific publications. Proteins with greater AlphaFold2 coverage exhibit substantially higher scientific attention after 2021, with publication probability rising by 60–80 percent and publication counts more than doubling relative to earlier years. Rather than substituting for experiments, AlphaFold2 appears to complement wet-lab research: experimental validation increases among proteins with higher coverage, while publication activity rises most for proteins with partial prior structural information. Teams studying high-coverage proteins tend to include more specialized and computationally oriented contributors, and their publications more frequently engage with structural and computational questions while maintaining experimental inquiry. Taken together, these patterns suggest that the introduction of a general-purpose AI tool may broaden the scientific opportunity frontier, not by displacing experimentation but by increasing its productive scope and shifting the allocation of human capital within the research system.