Google I/O 2026: Google wants AI to help scientists test ideas and analyse research papers

As artificial intelligence continues to spread across nearly every corner of the tech industry, concerns around automation, misinformation and job disruption remain difficult to ignore. Yet amid the flood of AI announcements at Google I/O 2026, Google also showcased a use case that could have meaningful implications beyond consumer apps and chatbots: scientific research.

The company introduced Gemini for Science, a collection of experimental AI tools designed to assist researchers with some of the most time-consuming and complex parts of scientific discovery. From generating hypotheses to analysing dense academic literature and conducting large-scale computational testing, Google says the new platform is intended to help scientists move from idea to insight more quickly.

More from Tech

The tools are initially experimental, but Google believes they could significantly reduce the manual effort associated with modern research workflows, especially in fields where scientists must process massive volumes of data and publications.

Gemini for Science: What does it do?

Gemini for Science consists of three primary tools: Hypothesis Generation, Computational Discovery and Literature Insights.

Hypothesis Generation is designed to help researchers during the earliest stages of scientific exploration. According to Google, the system scans millions of scientific papers, datasets and research findings to identify potential patterns, unanswered questions or promising directions for investigation.

The company said the tool does not merely produce speculative suggestions. Instead, the generated hypotheses are backed by “deeply verified” information and clickable citations, allowing researchers to trace the supporting evidence directly to its source.

The idea is to help scientists spend less time manually reviewing huge volumes of literature before identifying possible research pathways. In areas such as biology, chemistry and medicine, where new studies are published at overwhelming speed, this could prove especially valuable.

Once a hypothesis is created, researchers can move to Computational Discovery, which Google describes as an “agentic search engine” for scientific experimentation.

The tool can reportedly generate and evaluate thousands of experimental possibilities or test conditions far faster than traditional manual methods. Rather than individually designing each computational experiment, scientists can use the system to automate large parts of the discovery process and rapidly narrow down promising outcomes.

Google suggested that this capability could accelerate workflows in areas such as drug discovery, materials science and molecular research, where computational simulations often require extensive time and resources.

Literature Insights and wider availability

The third component, Literature Insights, focuses on helping researchers digest complex scientific information more efficiently.

The AI-powered assistant can search through scientific literature and generate simplified outputs such as reports, summaries, infographics and even audio or video explainers. Instead of reading through lengthy research papers one by one, scientists could potentially use the tool to understand developments across multiple studies more quickly.

As part of the broader initiative, Google also introduced Science Skills, a companion feature that connects with more than 30 major life science databases and research tools. According to the company, it can automate complex workflows that would normally take researchers hours to complete manually.

Google says access to Gemini for Science is beginning to roll out gradually starting today. Researchers interested in testing the experimental platform can apply through the Google Labs website, while enterprise organisations will also be able to access the tools through Google Cloud.

While questions around AI’s broader societal impact continue to dominate conversations, Google is positioning Gemini for Science as an example of how AI could be used not just to automate tasks, but potentially to accelerate scientific breakthroughs themselves.