Intended use:
Google reserves the right to update this intended use from time to time.
Last modified: 28 April 2026
Science experiments are intended to be used by scientific researchers to assist in the synthesis of scientific literature, development of hypotheses and research plans, and the optimisation of algorithms and models. These experiments can analyse publicly available scientific literature and research data. They leverage the Gemini models to: summarise relevant literature; populate structured data tables; identify research gaps and contradictions; propose testable hypotheses; suggest experimental approaches; and generate candidate algorithm code and predictive models tailored to user-defined scorable metrics. Much like working with a scientific brainstorming partner to generate ideas and computational tools that require further evaluation, Science experiments are designed to uncover new knowledge and formulate novel research proposals by building on existing evidence. It is not intended to automate the entire scientific or research process, and its output should be considered a starting point for research and further validation.
Science experiments are suitable for use by scientific researchers who are planning downstream experiments or developing computational methods. Although it has initially been validated most thoroughly by life sciences researchers, it is a general system that can be leveraged across disciplines, in research areas such as physics, chemistry, computer science and more. The outputs generated by Science experiments – including synthesised reports, generated ideas, code, algorithms and models – are suggestions for general experimental or computational approaches, and are not intended to be used without prior active review, validation, verification, meaningful modification and adaptation by the user. Users must independently verify all claims and data points against the source citations provided and defensible evidence. Science experiments and their output are not intended to be relied upon in a way that could expose humans to risk, including through any human subject research. The user is responsible for all onward testing of any hypotheses or algorithms generated by Science experiments. Additionally, Science experiments (including their outputs) are not intended, validated or approved for clinical use and should not be relied upon for individualised medical or other professional advice. Any output from Science experiments regarding these topics is for informational purposes only.