Google launches Gemini for Science to accelerate scientific research with AI

Google is betting that artificial intelligence can solve one of science’s biggest problems: information overload. As millions of research papers are published annually, individual scientists struggle to keep up with the flood of knowledge that could inform their work.

The tech giant announced Gemini for Science, a suite of AI-powered tools designed to help researchers handle complex scientific tasks that currently take weeks or months to complete manually. The initiative includes both experimental prototypes available through Google Labs and enterprise solutions for organizations.

Three experimental tools tackle research bottlenecks

Google Labs is rolling out three primary prototypes that target different stages of the scientific process:

• Hypothesis Generation: Uses a multi-agent system to simulate scientific debates, generating and evaluating research hypotheses with full citations for verification
• Computational Discovery: Runs thousands of code variations in parallel to test novel approaches in fields like solar forecasting and epidemiology
• Literature Insights: Searches and structures scientific literature into searchable tables, helping researchers identify gaps and create reports, slides, and other materials

The tools address a fundamental challenge in modern science: the gap between the exponential growth of scientific knowledge and researchers’ ability to process it all. “Scientific breakthroughs often rely upon making creative connections between data, but the time required to do this manually can take weeks or even months,” the company explained.

Science Skills brings specialized databases to researchers’ desktops

Beyond the experimental tools, Google is launching Science Skills, a specialized bundle that connects researchers to over 30 major life science databases including UniProt, AlphaFold Database, and InterPro. The system works through Google’s Antigravity platform, allowing complex analyses that normally take hours to be completed in minutes.

Google’s research teams have already tested the system on real scientific challenges. In one early test, researchers used Science Skills to perform complex analysis of a rare genetic disease caused by mutations in the AK2 gene, uncovering new insights in a fraction of the typical time required.

Enterprise adoption shows real-world impact

The enterprise versions of these tools are already being used by major organizations in private preview:

• BASF is using AlphaEvolve to optimize supply chains
• Klarna is applying the technology to enhance machine learning models
• Pharmaceutical companies like Daiichi Sankyo and Bayer Crop Science are using Co-Scientist for research acceleration
• U.S. National Labs are leveraging the tools as part of the Department of Energy’s Genesis Mission

Two research papers describing the ERA and Co-Scientist systems were published in Nature, providing academic validation for the approach.

Building trust through scientific collaboration

Google is working with over 100 institutions to validate its AI systems, including Stanford University on liver fibrosis research and Imperial College London on antimicrobial resistance. The company has assembled a “trusted tester community” ranging from PhD students to Nobel laureates to stress-test the systems.

The initiative also includes partnerships with major scientific conferences like ICML, STOC, and NeurIPS to develop tools for peer review and scientific validation.

Why this matters for scientific progress

The launch represents a significant shift toward using general AI agents rather than narrow, specialized models to accelerate scientific research. This approach could help address the growing challenge of scientific information overload while maintaining the rigor required for breakthrough discoveries.

These tools build on Google’s existing scientific AI successes, including AlphaFold, which has helped over 3 million researchers work on challenges from malaria vaccines to plastic-eating enzymes. The new Gemini for Science suite represents the company’s most comprehensive effort yet to integrate AI throughout the entire research process.

Researchers can register their interest in the experimental tools at labs.google/science, with access being gradually expanded over the coming months.