Google pulls Nano Banana image generator from Google Earth after misuse
Google Earth briefly shipped an AI image generator, built on Nano Banana (and Nano Banana 2), that turned a text prompt into a satellite, aerial or 3D flyover image. Google's blog post pitched it as a way to visualize historical sites and real estate projects, and the feature launched globally on the web version of Google Earth earlier this week. Henk van Ess of Digital Digging tested it and generated convincing fake scenes, including images depicting "refugees near the Mexican border" and a bomb crater near a hospital in Gaza. Some outputs were obviously fake, such as an image merging the Sphinx with the Statue of Liberty, but others were realistic enough to pass as genuine flyover photography. Google's initial response defended the feature's safeguards: "We take misinformation seriously, every image created with Nano Banana in Google Earth includes the SynthID digital watermark, so if someone is unsure about an image, they can ask the Gemini app or use Lens in Search to see if the image was AI-generated." Digital Digging tested those safeguards directly and reported fooling Hive's AI detector with an AI-altered video made from Google Earth footage. Van Ess suggests checking suspicious images by comparing them against other satellite platforms, such as Sentinel-2 or Landsat, and by verifying orbital data, including when an image claims to have been taken and by which satellite. On July 31st, after the story published, Google announced it is rolling back the feature. The source does not give an official reason from Google for the rollback, does not state a year for any of the events, and does not say how widely the fabricated images circulated beyond the specific examples named.
Key facts
- Google Earth's AI image generator, built on Nano Banana and Nano Banana 2, let users create satellite, aerial and 3D imagery from a text prompt.
- Henk van Ess of Digital Digging generated convincing fake images, including ones depicting refugees near the Mexican border and a bomb crater near a hospital in Gaza.
- Google said every image carries a SynthID watermark checkable via the Gemini app or Lens in Search, but Digital Digging fooled Hive's AI detector with an AI-altered video from the feature.
- Google announced on July 31st, after the story was published, that it is rolling back the feature.
Why it matters
An AI tool that can generate believable satellite and aerial imagery on demand undermines a category of evidence people treat as hard to fake. Photos and video already face doubt in the AI era; satellite and flyover imagery had remained a comparatively trusted format because faking it required specialist skill. A text-prompt tool built into a mainstream product like Google Earth removes that barrier, and Google rolled the feature back within days of it going global.
Who it affects
Journalists, fact-checkers and researchers who rely on satellite imagery as evidence are directly affected, since fabricated flyover images can now be produced without technical expertise. It also affects Google, whose watermarking and safety claims were tested and partly defeated in public before the rollback. Anyone who encounters a satellite or aerial image online now has an added reason to verify it rather than trust it at face value.
How to use it
The feature is being pulled, so there is nothing to use going forward. For images already generated with it, Google says every one carries a SynthID digital watermark that can be checked through the Gemini app or Lens in Search. Van Ess recommends additional checks: compare a suspicious image against other satellite platforms such as Sentinel-2 or Landsat, and verify the orbital data an image claims, including capture time and which satellite supposedly took it.
How solid is it
The account comes from a named researcher, Henk van Ess of Digital Digging, who produced and published specific example images and a video, and it includes Google's own quoted response and its follow-up rollback announcement on July 31st. The source does not state an official reason from Google for the rollback, does not give a year for the events described, and does not quantify how far the fabricated images spread beyond the examples cited.
Risks and caveats
Google's watermark may be hard to remove but is not impossible to miss or bypass: Digital Digging reported fooling Hive's AI detector with an AI-altered video generated from Google Earth footage. That means detection tools already available to the public do not reliably catch this kind of fabrication, so the rollback removes the tool but does not retroactively address images that may already be circulating.
“We take misinformation seriously, every image created with Nano Banana in Google Earth includes the SynthID digital watermark, so if someone is unsure about an image, they can ask the Gemini app or use Lens in Search to see if the image was AI-generated.”