Google has rolled back a newly launched AI image generation feature in Google Earth after researchers showed it could be used to create convincing fake satellite imagery on real-world coordinates. The feature, introduced on Friday and removed a day later, relied on the company's Nano Banana 2 model to let anyone zoom into a location, enter a text prompt, and produce a photorealistic aerial image layered over authentic map data.
Within hours, investigators and journalists demonstrated how easily the tool could be misused. Dutch intelligence researcher Henk van Ess published a blog post titled 'How to plant a nuclear plant in Iran,' showing that a single sentence could place a nuclear facility inside Iranian territory. He also generated an image of refugees on a road near the Mexican border and a hospital with a bomb crater in Gaza. Each fabricated scene was rendered directly on top of real satellite imagery and looked indistinguishable from the underlying terrain at a glance.
Fake imagery on real coordinates
The problem was not abstract. BBC Verify produced images of a toppled Eiffel Tower, a sinkhole swallowing the Great Pyramid of Giza, and Russian tanks in Kyiv. Agence France-Presse generated explosions in Paris and an Islamic State training camp in Syria. The technology publication 404 Media added skyscrapers to rural land and inserted a homeless encampment into Los Angeles. In every case, the output was geolocated to a real place and presented in the familiar visual language of commercial satellite imagery.
Van Ess noted in his analysis that it takes a government request to remove a single real building from Google Earth, but only a sentence to add a fake one. That asymmetry highlights why the feature alarmed people who work in verification, open source intelligence, and journalism. Satellite images are often treated as ground truth in conflict zones, disaster response, and environmental monitoring. A tool that can place a fabricated scene on a real coordinate weakens the trust that underpins that work.
A verification tool suddenly becomes a source of doubt
For years, Google Earth has been used by journalists and investigators to confirm the location of mass graves, missile strikes, refugee camps, and deforestation. The platform's value depends on the assumption that the imagery reflects what is actually on the ground. By allowing users to overlay AI-generated content on those same coordinates, Google created a situation where a single image could no longer be trusted without additional validation.
The timing adds to the concern. Misinformation tends to spread fastest during crises, when social media is flooded with unverified images and videos. In a fast-moving conflict, a fabricated satellite image showing destroyed buildings or military movements could go viral long before fact-checkers confirm its origin. Watermarking does not help in that window because the average viewer does not know to look for it.
Google's response and the rollback
Google acknowledged the issue in a post on its media relations account on X. The company said it had seen geospatial professionals using the feature for useful purposes, but it had decided to roll it back so it could implement stronger guardrails. Google also said none of the AI-generated images appeared in the main Google Earth platform