Google removed a new generative-image feature from Google Earth one day after releasing it. The tool combined Earth imagery with the company's Nano Banana 2 model so users could create custom scenes grounded in real locations.
Google presented the feature as a creative and professional aid. Suggested uses included reconstructing historical places for lessons, visualizing a building before construction, planning a real-estate project, and imagining how a neighborhood might change.
A trusted map meets synthetic imagery
The launch immediately created a credibility problem. Google Earth is widely used by journalists, researchers, investigators, and the public as a visual reference. Critics showed that screenshots of fabricated scenes could be separated from their original context and circulated as evidence of events that never happened.
Google said generated images were watermarked and did not become part of the shared Google Earth map. That reduced the chance of the application itself being permanently altered, but it did not prevent users from exporting or reposting convincing images.
In an update, Google said geospatial professionals had found useful applications for the feature. It also acknowledged that people were sharing generated imagery that appeared to violate company policies. Google rolled the feature back while it works on stronger guardrails.
The short-lived launch highlights a difficult product boundary. Image generators can already alter any screenshot, but placing generation inside a trusted geographic tool reduces the effort and can lend the result extra credibility. Labels must survive copying and sharing if they are expected to provide meaningful context.
Google has not provided a date for the feature's return. A revised version will need clearer visual separation between observation and imagination, along with controls that remain visible outside the original application.
