Gemini for Science: AI Tools for Scientific Discovery

Google DeepMind has launched Gemini for Science, a collection of AI-powered tools and experiments designed to accelerate the scientific method. By transitioning from narrow, specialized models to general agents, this suite aims to eliminate bottlenecks in data synthesis and hypothesis testing, allowing researchers to focus on high-impact scientific problems.

Experimental Tools for the Scientific Method

Google Labs is introducing three primary prototype tools to handle complex research tasks:

Hypothesis Generation (powered by Co-Scientist)

This tool simulates the scientific method by collaborating with researchers to define challenges and utilizing a multi-agent "idea tournament" to generate, debate, and evaluate hypotheses. To maintain scientific rigor, all claims are supported by clickable citations.

Computational Discovery (powered by AlphaEvolve and ERA)

Computational Discovery acts as an agentic research engine that generates and scores thousands of code variations in parallel. This allows scientists to test novel modeling approaches in fields such as epidemiology and solar forecasting at a scale and speed that would otherwise take months to perform manually.

Literature Insights (powered by NotebookLM)

Literature Insights allows researchers to search scientific literature and organize results into tables with custom, searchable attributes for side-by-side analysis. It enables the synthesis of findings across multiple papers, the identification of research gaps, and the creation of high-fidelity artifacts including reports, slide decks, infographics, and audio/video overviews.

Science Skills and Google Antigravity

Google has launched Science Skills, a specialized bundle that integrates data from over 30 major life science databases and tools, including the AlphaFold Database, UniProt, InterPro, and the AlphaGenome API.

When used on agentic platforms like Google Antigravity, these skills enable researchers to complete complex workflows—such as genomic analyses and structural bioinformatics—in minutes rather than hours. In early testing, this integration helped researchers identify potential mechanisms for a rare genetic disease caused by mutations in the AK2 gene.

Enterprise Integration and Real-World Impact

Gemini for Science capabilities are being extended to enterprise organizations via Google Cloud. Current private preview partners include:

• BASF: Utilizing AlphaEvolve to optimize supply chains.
• Klarna: Using AlphaEvolve to enhance machine learning models.
• Daiichi Sankyo, Bayer Crop Science, and U.S. National Labs: Utilizing Co-Scientist to tackle fundamental scientific challenges as part of the U.S. Department of Energy's Genesis Mission.

Ecosystem Collaboration and Validation

Google DeepMind is collaborating with over 100 institutions to validate these tools, including Imperial College London (antimicrobial resistance), Stanford University (liver fibrosis), and The Crick Institute.

To ensure the integrity of AI-generated insights, Google has established a trusted tester community of PhD students, industry researchers, and Nobel laureates. Additionally, the lab is piloting tools for agentic peer review and scientific validation, such as ScholarPeer and the Paper Assistant Tool (PAT), in collaboration with conferences like NeurIPS, STOC, and ICML.

Foundational Research and Publications

The release of Gemini for Science builds upon existing AI advancements such as AlphaFold and AlphaGenome. Validation for the new tools has been established through research papers on ERA and Co-Scientist, both of which were published in Nature on May 19, 2026.