What Happened
On July 30, Google added its Nano Banana 2 image generation model to Google Earth. You could zoom to any spot on the planet, type a sentence, and get an AI-generated image pinned to that location. Within 24 hours, Google pulled the feature entirely. The reason: users immediately started creating fake satellite imagery that looked disturbingly real.
Digital misinformation expert Henk van Ess typed one sentence into Google Earth and generated a nuclear plant in Iran. Then fake refugee streams at the Mexico border. Then a hospital and bomb crater in Gaza. All from a single text prompt, all placed on a platform the world trusts for accurate geographic data.
Google’s statement: “We know that people uniquely trust Google Earth for a reliable view of the world. We’ve seen people sharing screenshots of generated imagery that appear to violate our policies. We’re rolling back this feature while we work on implementing stronger guardrails.”
Why This Matters
Google Earth spent 20 years becoming the reference the world checks against. It is not just another app with an image generator bolted on. It is the place people go to verify what a location actually looks like. Adding AI generation to that platform, without the safeguards to prevent fake satellite imagery, was always going to be a problem. The fact that it took less than 24 hours for people to prove it is the story.
The images people generated were not subtle. Planes crashing into Manhattan skyscrapers. The Eiffel Tower toppled over. Blast craters in residential areas. These were shared as screenshots, stripped of the AI watermark Google said would protect against misuse. The SynthID watermark embedded in generated images only works if someone bothers to check it. Most people do not.
Google said the generated images “didn’t appear in the main Google Earth experience for others to see” and were watermarked. That defense misses the point. Screenshots do not carry watermarks. A convincing fake satellite image of a crisis zone shared on social media does not come with a verification button.
What the Tool Actually Does
Before the rollback, here is what Nano Banana 2 in Google Earth could do:
- Location-aware generation. You zoom to any coordinates on Earth and type a description. The model generates an image that matches the terrain, lighting, and visual style of that location. The geography is accurate. The content is fabricated.
- Photorealistic output. Nano Banana 2 produces high-quality images that blend seamlessly with actual satellite and aerial photography. The AI-generated content looks like it belongs on the map.
- One-prompt operation. No skill required. One sentence, one click. The barrier to creating convincing fake geographic imagery went from “photo manipulation expertise” to “type what you want to see.”
- SynthID watermarking. Every generated image included a digital watermark. But the BBC found it was possible to trick Gemini into saying the fake images were real, and community notes pointed out that SynthID only watermarks sections of generated media, leaving gaps that could be circumvented.
What to Look Out For
Google says Nano Banana will return to Google Earth once stronger guardrails are in place. When it does, the creative potential is real. Urban planners visualizing proposed developments. Filmmakers creating location scouts. Architects showing what a building could look like in its actual environment. The underlying technology is impressive. The deployment was the problem.
This is the tension every platform with AI generation is going to face. The better the output gets, the more responsibility comes with putting it in front of people. Google Earth is a trust platform. The same tool on a creative canvas like Gemini or Canva would generate excitement. On Google Earth, it generated fake evidence.
For anyone working with AI-generated imagery, this is the clearest example yet of why context matters. The same model that makes beautiful marketing visuals can make convincing misinformation when placed on a platform people trust for truth.
Watch for when Nano Banana returns to Google Earth. The guardrails Google puts in place will set the standard for every other platform that wants to mix AI generation with trusted data. If they get it right, it could be the most powerful creative tool on a map. If they get it wrong again, the consequences are obvious.
Want more like this? Join MOKU Club for free. Weekly resources, early access to new guides, and occasional templates you can actually use. Join below.
