Google Earth Now Generates AI Images Embedded in Real Satellite Photos

Google Earth is no longer just a map to look at, it's becoming a tool for visualizing things that don't exist yet. Starting July 30, 2026, anyone using the web version can pick any point on the globe, describe an idea in a plain sentence, and get back an image that looks like an actual photo of that specific place, just altered according to the prompt.

The mechanism is simple to use but technically nontrivial. Users zoom in on a chosen location, click the image-creation button, and type a description in natural language. The system doesn't generate an image from scratch on a blank background, instead it uses the real terrain geometry, street layout, buildings, and surroundings from Google Earth's database as its starting point.

How It Works in Practice

The team behind the feature stresses that what sets it apart from typical image generators is exactly this anchoring in real data. Instead of typing a prompt into a blank box and hoping the model understands the context, users start from an actual place on the map, and the AI fits the generated scene to its real perspective and spatial layout.

Rather than typing a prompt into a blank text box and hoping the AI understands what you mean, you start with an actual place on the map - Google Earth team

In practice, this means a request for a cyberpunk version of Google's Mountain View campus returns a futuristic city, but built on the same street layout and buildings that actually exist. In its promotional materials, Google shows examples such as a reconstruction of what Pompeii looked like before Mount Vesuvius erupted in 79 CE, and an infographic of the Statue of Liberty with automatically generated historical facts.

A Polish Test on Łódź and Warsaw

Journalists testing the feature also tried out less serious use cases. One prompt simply asked to show a given city, but prettier; another asked for a vision of the same place two hundred years from now in a cyberpunk style. The results of such queries show that the tool handles a playful makeover of a familiar neighborhood just as well as a serious urban-planning visualization.

The practical applications Google points to include visualizing changes to urban space, such as adding greenery and wider sidewalks to a given street, designing rooftop restaurants on existing buildings, or preparing educational infographics for historic sites. Architects and developers can use it to quickly show a client a project concept set within the real neighborhood, instead of waiting for a professional rendering.

Limitations and Caveats

Google explicitly notes that the tool remains experimental. The generated images are conceptual mockups, not actual construction plans or design documentation, so they can contain errors, geometry inaccuracies, and the hallucinations typical of generative models. The company doesn't recommend treating the results as a basis for actual investment or architectural decisions without review by specialists.

The limitations also extend to availability itself. The feature currently works only in the browser version of Google Earth, even though the mobile app has a far larger user base. Google AI plan subscribers have a daily limit on the number of images they can generate, and creating a single image can take up to two minutes. Changing the perspective mid-edit requires interrupting the process and resetting the view from scratch.

Part of a Broader Google Earth AI Strategy

The new image feature is part of a larger package of changes to Google Earth AI, which also includes Geospatial Reasoning, a Gemini-based framework that automatically combines different models analyzing satellite, population, and environmental data to answer complex questions. Google is also making some of these models available to business customers through Google Cloud, as well as to trusted testers in a Professional version in the United States.

For readers in Poland, the key point is that the image-generation feature is launching globally right away, without the regional restrictions typical of earlier Google rollouts in Europe. That means local urban planners, history teachers, or content creators can use it from day one, without waiting for a separate EU rollout.