Google releases AI tools to accelerate scientific research
Google is introducing Gemini for Science, a collection of AI tools designed to help researchers across disciplines move faster through routine tasks. Three experimental prototypes are now available in limited access, with enterprise versions already in use at companies and research institutions.
The core problem these tools address is straightforward: modern science generates too much data for individual researchers to synthesize manually. Scientists spend weeks or months making connections across papers, testing computational variations, and organizing literature. AI can handle these bottlenecks, freeing researchers to focus on identifying high-impact problems.
Three tools for different research stages
Hypothesis Generation uses a multi-agent system to help researchers define a challenge, then generates and evaluates competing hypotheses. The tool debates ideas and backs claims with citations to published papers.
Computational Discovery generates and scores thousands of code variations in parallel. This lets scientists test novel approaches to problems like solar forecasting or disease modeling that would normally take months to explore manually.
Literature Insights searches scientific papers and structures results into searchable tables for side-by-side comparison. Researchers can chat with the tool to uncover details, then generate reports, slide decks, infographics, or audio summaries from curated papers.
Google is accepting registrations for limited access to these experiments at labs.google/science.
Enterprise adoption already underway
Enterprise versions are in private preview with multiple organizations. BASF uses AlphaEvolve to optimize supply chains. Klarna uses it to improve machine learning models. Daiichi Sankyo, Bayer Crop Science, and U.S. Department of Energy labs use Co-Scientist to accelerate research on fundamental challenges.
Research papers validating these tools published today in Nature include work on both ERA and Co-Scientist.
Specialized tools for life sciences
Google is also launching Science Skills, a bundle that integrates data from over 30 life science databases including UniProt, AlphaFold Database, and InterPro. Used on agentic platforms like Google Antigravity, the tools perform complex workflows like structural bioinformatics in minutes instead of hours.
In early testing, researchers used Science Skills to complete an analysis that normally takes hours in minutes. The work uncovered potential mechanisms for a rare genetic disease caused by AK2 mutations.
Academic validation at scale
Google is collaborating with over 100 institutions to test the systems. Partners include Stanford University on liver fibrosis research and Imperial College London on antimicrobial resistance. A trusted tester community ranging from PhD students to Nobel laureates stress-tests the tools against real-world problems.
Google is also piloting agentic peer review tools with major scientific conferences including ICML, STOC, and NeurIPS.
Building on existing AI tools
AlphaFold, released earlier, has been used by over 3 million researchers on projects ranging from malaria vaccines to plastic-eating enzymes. AlphaGenome helps scientists identify disease drivers. These sit alongside existing research tools like Google Scholar, Earth Engine, and Colab.
For researchers looking to develop expertise in AI applications for science, AI for Science & Research courses cover research automation and scientific discovery workflows.
Your membership also unlocks:
