HeadlinesBriefing HeadlinesBriefing.com

Earth AI: Geospatial Agents for Planetary Challenges

Google AI Blog •
×

Google Earth AI, introduced earlier with a suite of geospatial foundation models, has evolved its Geospatial Reasoning agent into a far more capable experimental system. The agent is built for multi-step reasoning across planet-scale data and for autonomously training custom predictive models.

Google is sharing three updates. A new technical report presents systematic evaluations across four benchmarks, including two new ones, Geo KTools and Geo Expert Queries, which are being released so researchers can test their own geospatial agents. The results quantify the value of Earth AI's specialized tools and planning over out-of-the-box LLMs. Real-world partner testing has also begun, with a small group of social impact and commercial organizations, including the World Health Organization (WHO) Regional Office for Africa, People in Need, Give Directly, the Milken Institute and WPP, using the agent for disease outbreak investigations, disaster preparedness and clean energy screening.

Google is also expanding early access. Users can try an experimental deep research update for Ask Google Earth, and developers can use an agent-to-agent API with integration recipes for Gemini. The agent connects Gemini to Earth AI foundation models, including Weather Next, Flood Forecasting and the Population Dynamics Foundation Model (PDFM), along with Google Maps Platform, Google Earth Engine and Data Commons. Its capabilities are packaged as modular skills, and it builds a step-by-step plan before running complex operations. Every response shows that plan and the data sources used, so experts can verify assumptions.

Organizations can also bring their own data as CSV files, GeoJSON boundaries or cloud tables and join it with Earth AI's global models. The agent additionally powers the Planetary Prediction Engine, which takes local outcome data such as regional disease rates, crop yields or grid outages and automatically selects relevant population and environmental signals to generate predictions.

Source: Google AI Blog · Summarized by HeadlinesBriefing