As the climate crisis becomes increasingly urgent, humanity has finally gained an AI brain that can "read the Earth." On October 26, 2025, Google officially launched a major upgrade to Google Earth AI, deeply integrating the Gemini large model, transforming this classic geographic tool from a static map platform into an "Earth intelligence hub" with active reasoning capabilities. From now on, scientists, emergency response teams, and even urban planners can simply ask a natural language question and receive disaster risk analysis that previously took weeks in just a few minutes.
This new capability, called "Geospatial Reasoning," allows Google Earth to not only display satellite images but also actively connect weather models, population distribution, terrain data, infrastructure networks, and environmental monitoring information for multidimensional analysis. For example, inputting "Which areas will be affected by this heavy rain?" the system will combine real-time rainfall forecasts with surface runoff models to predict flood inundation areas; then asking "Which communities are most vulnerable?" the AI will immediately overlay population density, elderly ratio, and medical facility distribution to accurately identify high-risk populations; if further asking "Could power grids or roads be interrupted?" it can also retrieve power network and transportation layers to generate a comprehensive impact assessment.
The breakthrough is supported by three professional foundation models driven by Gemini: the Imagery model can identify surface changes such as deforestation and drying rivers; the Population model tracks human movement and settlement characteristics; the Environment model monitors air quality, water pollution, and vegetation health in real time — for example, automatically detecting harmful algal blooms in lakes and issuing early warnings about drinking water safety crises. These models are trained on decades of Earth observation data, combined with Gemini's multimodal reasoning capabilities, enabling them to handle ambiguous instructions, reconcile conflicting information, and generate probabilistic predictive scenarios.
The actual performance is astonishing. Official data shows that cross-domain environmental analyses that once required expert teams for weeks can now be completed in minutes. X platform users have described it as "addictive like a game": one person completed a smart mapping of parking lots and convenience stores within 300 meters of Tokyo Tower in just one minute; rescue organizations used the tool to provide shelter path recommendations for 15 million people during California wildfires. Non-profit organization GiveDirectly has also used it to accurately locate flood-affected areas and efficiently distribute cash aid.
To meet professional needs, Google has deployed the core models of Earth AI to the Google Cloud platform. Research institutions, governments, and companies can upload private data (such as local sensor networks or historical disaster records) for integrated analysis with public data. The U.S. professional version test channel (Trusted Tester) is now open, starting at $75 per month, supporting SSO login and API integration; Google AI Pro/Ultra subscription users enjoy higher call limits. In the future, non-profit organizations will also receive free support through Google.org. All data transmissions are encrypted, ensuring sensitive information is not used for model training, safeguarding intellectual property rights.
In competition with platforms such as Microsoft Azure Earth and ESA Sentinel, Google has taken the lead by leveraging Gemini's flexible reasoning and end-to-end conversation experience. Analysts point out that geospatial AI is becoming a key infrastructure for climate action — expected to reduce emergency response times by more than 25%, directly supporting the United Nations' Sustainable Development Goals.
