Google Earth Goes Deepfake

Google just turned Google Earth into an AI image generator sitting on top of trusted satellite maps, and researchers say that could turbocharge convincing “fake evidence” from space.

Story Snapshot

Google Earth now lets anyone overlay AI-generated scenes on real satellite and 3D map data.
Google says the tool helps people plan spaces and visualize history, and adds hidden AI watermarks.
Researchers warn it could help spread fake war damage, disasters, or political chaos tied to real locations.
The fight over this feature shows how much power a single tech company now has over what we see as “reality.”

Google bakes AI image generation into satellite maps

Google has added a “create image” button inside Google Earth on the web that lets users type a prompt and generate a new scene over real satellite, aerial, and 3D imagery. Product materials and coverage say the feature uses Google’s Nano Banana 2 image model, part of its Gemini artificial intelligence family, to build concept images tied to the exact spot on screen. A user can zoom to a town, click the tool, and ask for redesigned parks, different buildings, or imagined historical scenes.

Google is pitching this update as a way for everyday users, city planners, and educators to “visualize history, create real estate plans and more.” Company blogs about Earth AI say Gemini can now quickly find objects and patterns in satellite imagery, cutting some planning tasks from days to minutes. The same push markets Earth AI as a serious tool for environmental monitoring and disaster response, not just a toy image generator, which makes its new fantasy layer more sensitive.

Hidden watermarks and safety rules, but big gaps remain

Facing alarm online, Google told reporters and users that every image created with Nano Banana in Google Earth includes its SynthID digital watermark. SynthID is an invisible marker placed inside AI-made media so that tools like the Gemini chatbot or Google Lens in Search can later confirm if Google generated the content. The company also says it blocks “harmful topics” and warns in its AI policy against using its systems for misinformation or misleading activity.

Independent tests suggest these guardrails may be far weaker than Google implies. One researcher reported that none of his prompts were refused, even when he asked for a nuclear plant in Iran or violent scenes over real locations. A context note seen under Google’s own public response says SynthID only watermarks the AI-generated portions of Nano Banana images in Google Earth, in less than a quarter of tested pixels, not the entire image. That means large areas of the picture may look like standard satellite imagery with no clear AI mark.

Researchers warn of cheap, convincing geospatial deepfakes

Open-source intelligence analysts, journalists, and human rights researchers are sounding the alarm because they already rely on Google Earth to verify wars, protests, and disasters in places they cannot reach. For years, satellite maps have helped expose mass graves, bombed hospitals, and burned villages when local officials or powerful interests denied the damage. Now, the same tool can be used to drop fake buildings, explosions, or damage onto real coordinates with just a few words.

Analysts warn this lowers the cost of creating “evidence” for any story someone wants to push. A bad actor could screenshot an Earth AI image, crop out visible Gemini branding, and spread it on social media as proof of a missile strike or secret base. Because SynthID is invisible, ordinary viewers will not see any warning. They would have to upload the image to Gemini or Lens and ask if it is fake, which almost nobody does in the middle of breaking news. One third-party detector even rated a tested fake satellite image as only 0.8% likely to be AI-generated.

Why this hits shared distrust of “the elites” and big tech

People across the political spectrum already question whether powerful institutions tell them the truth, from Washington agencies to global corporations. This feature lands in that climate. Google now controls the satellite platform, the AI model, and the moderation rules, which means one company decides both what is possible to fake and how those fakes are tagged. Researchers say there has been no independent audit of how well SynthID survives cropping, recompression, or reposting.

Re: our statement on Image Generation in Google Earth:
“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…

— News from Google (@NewsFromGoogle) July 31, 2026

For conservatives who worry about propaganda and “deep state” manipulation, the idea that anyone can paint new damage or infrastructure onto maps used by the media feels like an open door to abuse. For liberals who worry about growing inequality and abuse of power, it looks like another case where a giant tech firm ships a risky product without proving its safety, then asks the public to trust invisible fixes. Experts argue that scientists, reporters, and citizens will now have to do more manual cross-checks against other satellite sources to be sure what they are seeing is real.

What comes next: more testing, more pressure on Google

Researchers are calling for independent “red team” testing to see how easily the tool can be used to fake infrastructure, land use changes, disaster scenes, or political events. They also want clear public rules on what prompts are blocked and regular reports on abuse, including how often harmful prompts get through. Without that, they say, Google’s claim that its safeguards offset the risk of misinformation remains a promise, not a proven fact.

Some outlets report that Google has already had to pause or roll back AI image features after public backlash in the past, including over biased or historically inaccurate images of people. That history makes critics even less willing to take its word that this new satellite tool is “safe enough.” The larger question is how much trust citizens should place in digital maps at all, when the same system can now be used to show either honest damage or made-up chaos. Watching how Google responds to the growing pressure on Earth AI will be a key test of whether the public can still rely on its view of the world from above.