Google has shut down a new artificial intelligence feature in Google Earth that allowed users to generate custom images over real-world satellite, aerial, and 3D imagery. The company announced the removal of the tool on July 31, 2026, just one day after its initial rollout.
Google Rolls Back AI Satellite Imagery Tool Following Misinformation Concerns
The feature integrated Nano Banana 2, an AI image generator, into the Google Earth platform. It enabled users to enter text prompts to create concepts grounded in the real world, such as visualizing how the ruins of Pompeii might have appeared in 78 A.D. or imagining a community garden in an empty neighborhood lot.
In a statement released on X, Google acknowledged that while geospatial professionals used the feature for useful purposes, the company observed people sharing screenshots of generated imagery that appeared to violate its policies. Google stated it is rolling back the feature to work on implementing stronger guardrails,
noting that the generated images were watermarked and did not appear in the main Google Earth experience for other users to see.
High-Fidelity Deepfakes and Verification Risks
The tool’s ability to “weld” invented imagery to genuine coordinates drew immediate criticism from open-source investigators and forensics experts. Henk van Ess, a researcher with Digital Digging, warned that such forgeries inherit the credibility of the map they are born on. During testing, van Ess reported that the tool did not refuse prompts for a refugee camp on the US-Mexico border, a bomb crater next to a hospital in Gaza, or a nuclear plant in Iran.
Other organizations conducted similar tests to demonstrate the potential for abuse:
- NPR generated images of a flooded U.S. Capitol complex and multiple fires on Iran’s Kharg Island, a strategic oil terminal.
The Erosion of Satellite Imagery as a “Safe Bet”
For years, satellite imagery has served as a critical tool for journalists and investigators to verify events on the ground, particularly in hard-to-reach areas. Jake Godin, a senior researcher with Bellingcat, noted that satellite imagery has historically been a “safe bet” for verification because it is difficult to fake.
However, the introduction of high-fidelity AI generation threatens this reliability. Experts warn that the existence of such tools may allow governments to claim that real satellite images are fake, further undermining public trust.
The risk is not entirely new; during the U.S.-Israel war with Iran, fake images of a damaged U.S. base in Bahrain circulated in Iranian media. Those images reportedly used Google Earth and Google AI to generate fake damage, though real satellite imagery later confirmed the base was damaged in a different manner than shown in the fakes.
Technical Safeguards and Their Limitations
Google defended its initial launch by stating it takes misinformation seriously
and implemented several protections:
- SynthID Watermarking: All generated images were marked with a digital watermark to ensure they would be flagged as AI-generated in tools like Gemini.
- Topic Restrictions: The company claimed it prevented the creation of images on “harmful topics.”
- Verification Tools: Google suggested users unsure of an image could use the Lens feature or the Gemini chatbot.
Despite these measures, researchers found gaps in the protections. BBC Verify reported that it was possible to circumnavigate checks and trick Gemini into identifying fake Google Earth images as real. Additionally, Henk van Ess stated he was able to fool Hive’s AI detection tool with a video generated in Google Earth, asserting that during his testing, nothing was refused, nothing was softened.
