Google unveils experimental AI suite for scientific research at I/O 2026
The tech giant’s new tools, including Hypothesis Generation and Computational Discovery, are rolling out via Google Labs and Cloud partnerships, though their experimental status means real-world efficacy remains unverified.
Google has introduced Gemini for Science, a new collection of experimental artificial intelligence tools designed to streamline scientific research workflows. Announced during the Google I/O 2026 conference, the suite addresses the time-intensive nature of scientific discovery by automating key stages of the research process, from initial hypothesis formation to literature summarisation.
The core of the offering comprises three primary features: Hypothesis Generation, Computational Discovery, and Literature Insights. Hypothesis Generation utilises the model to analyse millions of scientific papers to propose theories or challenges. Google states that the tool ensures "absolute rigor" by providing claims that are deeply verified and supported by clickable citations, aiming to reduce the manual burden of initial theory development.
Following hypothesis creation, the Computational Discovery tool functions as an agentic search engine capable of generating thousands of tests and experiments. Google describes this capability as significantly faster than manual methods, allowing researchers to rapidly validate theories. The third component, Literature Insights, operates as an AI-powered chat interface that digests scientific literature into more accessible formats, including reports, infographics, and audio or video overviews.
In a separate release, Google is launching Science Skills, a tool designed to help researchers navigate complex workflows using data from over 30 life science databases. This feature aims to condense tasks that typically take hours into minutes, further targeting the inefficiencies inherent in manual data processing within the life sciences sector.
Access to these experimental tools is being rolled out gradually, commencing immediately. Researchers can request access through a form on the Google Labs website, while enterprise organisations will be able to utilise the suite through partnerships with Google Cloud. As these tools are classified as experimental, their long-term reliability and performance in real-world scientific settings have not yet been independently verified.
