Google's Gemini for Science wants to cut research workflows from hours to minutes

Google has launched Gemini for Science, a suite of AI research tools announced at Google I/O 2026. The package bundles three distinct tools aimed at different stages of the scientific process — from forming a hypothesis to sifting through mountains of published literature. Early enterprise partners including Bayer Crop Science and Daiichi Sankyo are already in private preview, suggesting real-world pharmaceutical and life-science teams see genuine potential here.

The tools

Hypothesis Generation scans millions of scientific papers and surfaces new research directions or testable hypotheses. Critically for academic use, every result comes with verified, clickable citations — so researchers can trace claims back to primary sources rather than taking the AI's word for it.

Once a hypothesis is in hand, Computational Discovery takes over. Google describes it as an "agentic search engine" that can auto-generate thousands of experimental variants far faster than manual methods allow. Think of it as an accelerated trial-and-error layer sitting between an idea and a lab bench.

The third tool, Literature Insights, is a chat interface for scientific literature. Feed it a topic and it produces summaries, infographics, and even audio or video overviews — useful for researchers trying to get up to speed in an adjacent field without reading hundreds of papers.

Alongside the three main tools, Google also announced Science Skills, a component that connects to more than 30 life-science databases — including UniProt, the AlphaFold Database, and InterPro — and claims to shrink complex multi-step workflows from several hours to a few minutes.

What's still unclear

Access is rolling out gradually. Individual researchers can apply via Google Labs, while enterprise clients get access through Google Cloud. No pricing has been announced for either tier.

Data residency and research integrity questions remain open. UK universities already require AI transparency declarations on submitted work, and it's not yet clear whether AI-generated hypotheses will satisfy peer-review standards that expect meaningful human oversight. Google hasn't confirmed specific data-handling guarantees for EU or UK jurisdictions, which matters for institutions bound by GDPR or the UK GDPR framework.

Early adopters like BASF and Klarna are testing AlphaEvolve in parallel, per the official Google blog. Whether the speed gains hold up under real research conditions — and whether journals will accept AI-assisted outputs — are the questions that will define how widely this actually gets used.