Nano Banana in Google Earth was a short-lived experiment that Google never should have launched in the first place. Last week, the tech giant rolled out its latest AI image generation model, Nano Banana 2, directly into its widely used world explorer program.
The company heavily marketed the integration as a creative tool. Users were encouraged to generate educational infographics, visualize historical landscapes over modern-day map imagery, or even project their “dream homes” onto real-world backdrops to plan future renovations. This move followed a massive, ongoing trend of Google aggressively shoehorning its AI platforms into nearly all its products, such as adding new Gemini capabilities to Workspace tools.
From a purely corporate perspective, merging their latest image generator with their mapping software probably made perfect sense. In reality, it proved to be a disastrous combination of technology and utility.
The Immediate Rollback
Just 24 hours after launching this massive integration, Google quietly pulled the plug. The company updated its original announcement with a retraction statement:
“We know that people uniquely trust Google Earth for a reliable view of the world. We’ve seen geospatial professionals using this feature for a range of useful purposes; however, we’ve also seen people sharing screenshots of generated imagery that appear to violate our policies. So we’re rolling back this feature in Google Earth while we work on implementing stronger guardrails. It’s important to note that generated images didn’t appear in the main Google Earth experience for others to see and were watermarked as AI-generated.”
While Google framed the rollback around “policy violations,” that mild description drastically underplays exactly what users were doing with the tool.
A Disinformation Machine Built on Trolls
Predictably, internet trolls immediately used Nano Banana 2 to generate highly insensitive, inappropriate, and offensive images pinned to real-world coordinates. Among the most viral generated images were a plane crashing into the new World Trade Center in Manhattan, the Eiffel Tower reduced to rubble, and the Egyptian pyramids collapsing into a massive sinkhole.
While most internet-savvy users recognized these images as dark jokes, the deeper threat was the potential for rapid, widespread misinformation. One user perfectly demonstrated the danger by using the AI to seamlessly place crowds of refugees shoulder-to-shoulder along the Mexican border. Because the AI generated these fabrications over actual satellite imagery, an unsuspecting viewer scrolling through social media could easily be tricked into thinking they were looking at a real, official Google Earth screenshot.
While Google likely had some baseline safety guardrails in place, they clearly were not robust enough. Because the tool was freely accessible to anyone who opened Google Earth, bad actors simply had to tweak their prompts until the AI complied with their malicious requests.
Why Generative AI Ruins Educational Tools
You do not need to be a staunch AI skeptic to see why putting a hallucinating image generator into a digital globe is a terrible idea. Google Earth is fundamentally an educational tool. Users log on to learn about geography, explore foreign cities, and trust that the Street View data provides an accurate, reliable look at reality.
Introducing a tool designed specifically to fabricate reality directly undermines the platform’s core purpose.
- The Illusion of History: Google touted the ability to view “historical landscapes,” but Nano Banana 2 was not showing users accurate history. If prompted to show Paris in 1726, the AI simply hallucinated generic period-appropriate buildings and overlaid them on the map. It offered an entertaining illusion, not an educational fact—essentially reducing Google Earth to the same level as AI-generated “history slop” on social media.
- Unvetted Infographics: The AI was also promoted as a way to generate educational infographics about specific locations. However, these graphics were entirely unvetted. Nano Banana 2 could easily spit out half-truths and hallucinations without citing any clickable sources to verify its claims. Casual users would inevitably retain this fabricated data as fact.
Ultimately, this is not an argument against generative AI as a whole, but rather a warning against embedding hallucination-prone tools into platforms built on factual reality. At a time when digital misinformation is already spreading at breakneck speeds, tech companies should be looking for ways to verify the truth—not providing the public with a built-in disinformation engine just to claim another AI “success.”
