Google is unleashing a suite of AI tools called Gemini for Science designed to tackle one of modern research's deepest paradoxes: we know more than ever, yet individual scientists can see less of the full picture. As humanity's collective knowledge grows exponentially, researchers face a mounting bottleneck—scientific breakthroughs often depend on making creative connections across vast bodies of work, yet synthesizing this manually can consume weeks or months of painstaking effort.
The company's answer is a collection of experimental prototypes that transform AI into a "force multiplier" for human ingenuity. Rather than replacing researchers, these tools are built to handle the time-consuming mechanical work, freeing scientists to focus on the creative and strategic thinking that drives real discovery.
Three core prototypes form the foundation of Gemini for Science. Hypothesis Generation, built with Co-Scientist, reimagines ideation itself. Scientists collaborate with the AI to define a research challenge, then the system runs what Google calls an "idea tournament"—generating, debating, and evaluating hypotheses across the mountains of published research no human could synthesize alone. Every claim is rigorously verified with clickable citations to ground results in evidence. Computational Discovery, combining AlphaEvolve and the Empirical Research Assistance tool, tackles another bottleneck: testing novel approaches. The system generates and scores thousands of code variations in parallel, compressing work that would take months into manageable timeframes. This opens possibilities in fields like solar forecasting and epidemiology where computational experiments are central to progress. Literature Insights, powered by Google NotebookLM, transforms how researchers navigate scientific papers. It searches the literature and structures results into searchable, sortable tables for side-by-side comparison. Researchers can then chat with their curated corpus to uncover nuances, then generate polished outputs—reports, slide decks, infographics, and even audio and video overviews—that synthesize findings and identify research gaps.
The tools aren't confined to academic laboratories. Companies are already driving real-world impact through enterprise versions. BASF is using AlphaEvolve to optimize supply chains. Klarna is leveraging it to enhance machine learning models. Pharmaceutical firm Daiichi Sankyo and agricultural biotech company Bayer Crop Science are using Co-Scientist to accelerate research, as is the U.S. National Labs through the Department of Energy's Genesis Mission. The validation is mounting: research papers on ERA and Co-Scientist publish today in Nature, the world's leading scientific journal.
Beyond these headline tools, Google is launching Science Skills—a specialized bundle that integrates insights from over 30 major life science databases and resources including UniProt, the AlphaFold Database, AlphaGenome API, and InterPro. Running these skills on agentic platforms like Google Antigravity allows researchers to complete complex analyses—structural bioinformatics, genomic work—in minutes rather than hours. Early testing has already validated this speedup in practice.
Beginning today, researchers can register interest at labs.google/science to gradually gain access to these experiments. Google Cloud is simultaneously expanding access to enterprise-grade versions for organizations ready to deploy these capabilities at scale. The shift reflects a fundamental belief: the next era of scientific discovery won't be powered by narrow, specialized models, but by general agents that empower researchers across every field to see further and move faster.