Google Earth has always been that tool we open when we want to confirm something with our own eyes — after all, it is a real satellite image, right?
Well, all that credibility nearly became an open door for misinformation on a global scale.
Less than 48 hours after launching a new artificial intelligence feature, Google had to take a step back and pull the plug after experts and the BBC Verify team showed, in practice, just how big the problem really was.
The tool allowed users to create fake images overlaid on real satellite imagery — and the results were downright alarming.
The Eiffel Tower destroyed, the Great Pyramid of Giza swallowed by a crater, Russian tanks rolling through the Ukrainian capital.
All of it generated with a few clicks, wrapped in the look and feel of a trusted source.
And when detection tools were put to the test to flag these images as fake, they did not always work — which raises a pretty serious question about the pace at which new AI features are being released into the world. 🤔
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
What Was This Feature, Anyway?
The feature was part of an integration that brought Google’s image generator called Nano Banana 2 directly into Google Earth. The idea was to let users generate scenarios on top of a base layer of real satellite, aerial, or 3D imagery. On paper, it sounds like a creative use of the technology. In practice, the distance between legitimate use and abuse turned out to be much shorter than expected.
All you had to do was type a text prompt — a simple instruction — and the AI would generate an image that visually fit right over the real photo of the chosen location. From there, the content could be downloaded and shared across other platforms. The final result retained the texture, angle, and resolution typical of Google Earth imagery, making the generated content extremely difficult to tell apart from the real thing at first glance. No advanced technical knowledge was needed, no professional editing software, no special skills. Anyone with access to the tool could produce a convincing image in a matter of minutes.
The core problem was exactly that combination: ease of use paired with the historical credibility that Google Earth carries. As Google itself acknowledged in its statement, people trust Google Earth in a unique way to get a reliable view of the world. When someone shares a screenshot with the recognizable platform interface, the natural tendency is to trust it. Using that same visual identity to wrap AI-fabricated content was a perfect recipe for spreading misinformation at massive speed and scale.
Why Is This So Dangerous?
The issue here goes far beyond funny images or creative experiments. The real risk lies in the context where this type of content can be used — and in the people who can be fooled by it. As AI and misinformation expert Henk van Ess explained to BBC Verify, Google allowed invented images to be welded to genuine coordinates, drawn over genuine imagery. According to him, the forgery does not need to look convincing on its own — it inherits the credibility of the map it was born on.
The cases documented during testing were pretty eye-opening. Van Ess created fake images of a nonexistent nuclear plant in Iran, a refugee camp on the border between the United States and Mexico, and even a fictional hospital in Gaza with a bomb crater next to it. Any one of those images, if published without context during a moment of political or social tension, would have real potential to spark reactions, fuel false narratives, and trigger concrete responses based on something that simply did not exist. That is the kind of impact visual misinformation can have when it is dressed up in a seemingly trustworthy source.
AI detection researcher Henry Ajder told BBC Verify that the creation of fake images in conflict zones can be especially destabilizing when events are unfolding quickly and information is scarce. He noted that journalists might be able to verify an image and report on it, but not everyone is going to do that verification work. For Ajder, spaces we assume are free from AI are valuable, especially for journalists and people working on the front lines of fast-moving crises. 😬
Did the Safeguards Actually Work?
Google initially stated that the guidelines built into its AI model would prevent the creation of images depicting harmful themes. But BBC Verify found that small changes in prompts were enough to bypass those protections. One revealing example: a request to create a raised platform for hanging traitors next to the UK parliament was rejected. But when a less specific prompt was used, the request went through and the image was generated anyway.
In another test, the team asked the tool to draw the word Trump on the White House lawn, and it refused based on its guidelines. However, it allowed a community garden to be generated in that exact same spot. These tests show how the safety barriers, while they did exist, were fragile enough to be worked around by anyone willing to rephrase a request.
Did Detection Tools Hold Up?
Google explained that content created by its AI tools contains invisible watermarks to indicate the image is fake or manipulated. The company also said that anyone with doubts about an image can check it using the Gemini chatbot app or the Lens feature in search.
Generally speaking, BBC Verify’s tests showed that this works with images generated by the new feature. But — and here is the concerning part — it was also possible to bypass those checks and trick Gemini into saying that fake Google Earth images were real. On top of that, tests with external detection tools in some cases completely failed to identify the AI-generated content.
It is worth understanding why these tools fall short. Most of them analyze statistical patterns in pixel structure, lighting inconsistencies, compression artifacts, and other technical signals that typically show up in synthetic images. The problem is that more modern artificial intelligence models are getting increasingly better at replicating exactly those details. And when generation happens on top of a real image base — as is the case with overlays on Google Earth — many of the signals that detection tools look for simply do not appear in the same way.
The current landscape works a bit like a cat-and-mouse game on a global scale. Every advance in generation tools demands a corresponding response from detection tools — and that response usually takes time. In the gap between the two, there is a window of vulnerability that can be exploited to spread fake images before any control mechanism can step in.
Why Is Trust in Satellite Imagery at Stake?
For a lot of people, Google Earth is the entry point for open-access satellite imagery and a way to get information about hard-to-reach areas. That was exactly the point that geospatial analyst Bill Greer, co-founder of the nonprofit satellite service Common Space, raised with BBC Verify. According to him, this is potentially very damaging to public trust, because satellite imagery has historically been viewed as an especially reliable data source.
Greer also pointed out that fake satellite imagery has been used to spread misinformation before. And he flagged a worrying trend: governments are increasingly restricting access to data, which means we will likely see more fake images and fewer real ones — something that will erode trust in this type of source even further.
Google Pulled the Feature — What Now?
Google’s decision to suspend the feature less than two days after launch was, on one hand, a positive sign that the company reacted quickly to pressure from experts and fact-checking outlets. On Friday, Google said it had seen people sharing screenshots of generated images that appeared to violate its policies, and so it paused the feature while working on implementing stronger safeguards.
On the other hand, the simple fact that the feature reached the public without these risks being identified internally raises questions about how review and safety processes are structured within major tech companies when it comes to artificial intelligence. The speed at which new products are being launched keeps increasing, and that race does not always go hand in hand with a careful assessment of the consequences.
The Google Earth episode fits into a broader pattern that keeps repeating across the industry: a new feature ships and the real problems only surface once a significant number of people start using the tool in ways that were not anticipated. That is not necessarily bad faith — it is genuinely hard to predict every possible use case for a new technology. But when the risks involve creating fake images with the potential to cause real harm, the acceptable margin for that kind of surprise is much smaller.
As Henry Ajder summed it up, the situation is concerning because it represents yet another place where trusted information sources are being eroded by the constant doubt that AI might be involved. What we should expect going forward is that Google and other companies in similar situations adopt more rigorous impact assessment processes before making AI content generation features available to the public. 🛰️
This episode serves as an important reminder: powerful technology demands equally powerful responsibility. And when a trusted platform’s name is on the line, the impact of a mistake can reach far beyond the digital world.
