Google debuts Gemini for Science tools

Google introduced Gemini for Science at its I/O conference on May 19, 2026, expanding its AI push into research workflows with a set of experimental tools for scientists. The company said the new collection is designed to help researchers generate hypotheses, run computational experiments and review scientific literature faster. Google also tied the launch to Google Antigravity, its agent-focused development platform, and to existing systems including Co-Scientist, AlphaEvolve, Empirical Research Assistance and NotebookLM. Google said access is opening gradually through Google Labs and for enterprise users through Google Cloud. ### What exactly did Google launch under Gemini for Science? Google’s May 19 blog post described Gemini for Science as “a collection of science tools and experiments” meant to expand the scale and precision of scientific exploration. The company said the initial rollout includes three main prototypes: Hypothesis Generation, Computational Discovery and Literature Insights. Engadget, citing Google’s announcement, said the tools are aimed at three core research tasks: generating ideas, testing them and understanding prior literature. (blog.google) Google said the tools were built with existing DeepMind and Google Research systems, including Co-Scientist for hypothesis work, AlphaEvolve and Empirical Research Assistance for computational testing, and NotebookLM for literature analysis. ### How is Google describing the first tool, Hypothesis Generation? (blog.google) Google said Hypothesis Generation is built to help researchers define a scientific challenge and then use a multi-agent “idea tournament” to generate, debate and evaluate possible hypotheses. The company said the tool is meant to address the volume of published research, which it said no individual scientist can fully synthesize manually. Engadget reported that Google presented the tool as a way to work through millions of scientific papers during the early stages of the scientific method. (blog.google) Google said the resulting claims are “deeply verified and supported by clickable citations,” a feature it said is intended to support rigor in the research process. ### What does Computational Discovery do? Google said Computational Discovery is a prototype research engine that generates and scores thousands of code variations in parallel. (blog.google) The company said the tool is intended to increase the number of hypotheses researchers can test with computational experiments. Google’s examples included modeling work in solar forecasting and epidemiology, which it said can take months to navigate manually. (engadget.com) ZDNET’s live coverage of I/O said Demis Hassabis presented Gemini for Science with use cases including digital twins of the Earth for climate and weather simulation and support for drug-discovery workflows. ### How does Literature Insights fit into the workflow? Google said Literature Insights searches scientific literature and structures results into tables with custom, searchable attributes for side-by-side analysis. (blog.google) The company said the feature is built with NotebookLM and is intended to help researchers compare papers and organize findings more quickly. Engadget reported that the tool can also generate more digestible outputs from scientific literature, including reports, infographics and audio or video overviews. (blog.google) That places it at the review stage of research, after idea generation and computational testing. ### What is Science Skills, and who gets access first? Google’s blog said Gemini for Science also includes new Science Skills in Google Antigravity. (blog.google) Engadget reported that Science Skills can pull from more than 30 major life-science databases and tools so researchers can complete workflows the company said would otherwise take hours. Google said on May 19 that the Gemini for Science tools are experimental and are being rolled out gradually starting that day. (engadget.com) The company directed interested users to Google Labs for access requests and said enterprise organizations will also be able to obtain the tools through Google Cloud. ### Where does this sit inside Google’s broader I/O push? Sundar Pichai said in his I/O keynote transcript that Google is in “the agentic Gemini era,” framing the conference around AI systems that can do more than assist with writing. (blog.google) Google’s I/O collection page grouped Gemini for Science alongside broader announcements including Gemini Omni, Gemini 3.5, AI Search updates and agentic tools across consumer and enterprise products. Google’s next step is the gradual release through Google Labs and Google Cloud announced on May 19. (engadget.com) Researchers seeking access can use Google’s signup process, while the company’s I/O 2026 collection page lists the related product updates and follow-on announcements from named executives including Sundar Pichai, Koray Kavukcuoglu, Pushmeet Kohli and Yossi Matias. (blog.google 1) (blog.google 2)