Google said at Cloud Next 2026 that it is adding three Gemini AI features to its Maps and Earth platforms, expanding their role beyond navigation into enterprise AI workflows. The update includes scene generation grounded in Street View, automated analysis of satellite imagery, and experimental access to Earth AI models trained to detect physical infrastructure.
Street View grounding enters private preview in the US
The first feature lets enterprise users type a prompt into the Gemini Enterprise Agent Platform and generate visual content inside real-world Street View environments. WPP is already testing the capability for immersive advertising work. For now, it is limited to US locations and remains in private preview.
The practical use case is straightforward: a brand can preview how an ad placement would look on an actual street corner, with real buildings and sidewalks in the background, instead of relying on a fully synthetic 3D mockup.
Satellite analysis moves from weeks to minutes
The second feature, described as aerial and satellite imagery insights, brings Google Earth imagery into BigQuery for automated analysis. Google said this can support tasks such as tracking residential construction progress or assessing building damage after disasters. According to the company, work that previously took weeks of manual image review can now be completed in minutes.
Earth AI models open to enterprise experimentation
The third release is a pair of Earth AI imagery models now available for experimental access in Google Cloud Model Garden. These models are trained to identify objects in satellite images, including bridges, roads, and power lines. In the past, companies trying to build similar tools often had to create and train their own AI systems, a process that could take months.
Google partner Vantor has already integrated the models into its disaster recovery application, Sentry, where they are used to automatically flag damaged infrastructure after extreme weather events.
Maps is being positioned as an AI perception layer
All three products share the same premise: geospatial data is no longer just an answer to location queries, but a perception layer for AI agents interacting with the physical world. Google had earlier released Maps Grounding Lite through MCP, allowing developers to connect LLMs to Google Maps’ 300 million places database. FIFA World Cup 2026 and the Boston Marathon are using that capability for AI-powered event guides, while TUI has used it to turn static itineraries into live personalized recommendations.
The same direction is visible on the consumer side. Ask Maps allows users to query nearby places conversationally using data from 500 million community contributors, and Gemini can analyze Street View and aerial imagery to generate 3D route guidance with real building facades. Taken together, the new releases show Google pushing Maps toward a broader role as infrastructure for AI systems that need to interpret the real world.