Here’s the problem with putting an AI image generator in Google Earth
Now a text prompt is all it takes to generate reality-warping images using Google Earth's satellite, aerial, and 3D imagery, like these images generated by Digital Digging's Henk van Ess that show "refugees near the Mexican border" and a bomb crater near a hospital in Gaza.
Google Earth now lets a user type a sentence and receive a photorealistic image built from its own satellite and aerial archives. The feature, powered by the Nano Banana model, can place refugees at the Mexican border or a bomb crater beside a Gaza hospital without ever leaving the application. The capability is not a glitch. It is the product.
Google's defense is procedural. Every generated frame carries a SynthID watermark, and the Gemini app or Google Lens can be consulted to confirm provenance. The framing assumes a curious, careful viewer. The actual audience for synthetic war photography is a newsroom on deadline, a propaganda account, or a courtroom exhibit, none of which pause to check a watermark before circulation.
The deeper issue is structural. Google Earth has long served as a reference layer, a place where journalists, NGOs, and litigants verify what a location looks like. By collapsing the line between capture and composition, the company has converted a verification tool into a fabrication surface. The watermark protects Google from the accusation of deception. It does not protect the public record from contamination.
The market signal is straightforward. Generative media is moving from green-field canvases into the trusted archives that anchor factual reporting. Each integration narrows the gap between what is real and what is rendered, and shifts the burden of proof onto the consumer of every image, map, and aerial view they encounter from now on.