Google Earth adds Nano Banana image generation on web

Google Earth just picked up an AI image generator. On July 30, Google announced that Nano Banana 2, its image model, is live inside Google Earth on the web: zoom to any place, click a button, describe what you want to see, and the model returns a picture generated against real satellite, aerial and 3D imagery.

The angle that makes this interesting is not that Google shipped another image model. It is that the model is being pointed at a specific structured data source, Earth's actual geospatial imagery. Google's blog, written by Google Earth product manager Bryan Horowitz, pitches use cases like rendering Pompeii as it appeared in 78 A.D., dropping a proposed lakefront cabin onto an empty lot, or reimagining Google's Mountain View campus as a futuristic sci-fi utopia. Forbes reports the model can 'draw from multiple sources of information to help generate its output', with worked examples including a Statue of Liberty infographic and Venice projected with future water levels.

For teachers, architects and real estate teams, this is a genuinely new fast-concepting surface: you can sketch a what-if for a specific location without leaving Google Earth. That part is real. What is less clear is where the boundary sits between grounded rendering and pure invention. A 'historical visualization' of an ancient city is, in the end, a plausible-looking guess, and a proposed development render is not a survey.

The honest caveats are that this is web-only at launch. Engadget notes mobile apps are not covered, and the reporting does not spell out the daily generation limit, the subscription tier that gates it, or the specific provenance markers the outputs carry. Those are the details that will decide whether this becomes a serious workflow tool or a novelty. If Google extends it to Earth's mobile install base and is explicit about the limits and watermarking, it becomes one of the more interesting practical uses of generative image models outside a chatbot window.