The tool that verifies war crimes, tracks deforestation, and confirms troop movements just became, for one day, a disinformation machine anyone could operate for free.

Open-source investigators built their workflows around satellite imagery they could trust

Verification researchers and journalists rely on Google Earth as a ground-truth reference when confirming the location of real-world events. A feature that lets anyone modify that imagery with AI does not add capability to that workflow — it destroys the premise of it entirely.

Google shipped a fake satellite image generator and called it a classroom tool

Google Earth risked ruin with retracted AI tool for making fake satellite pics briefly integrated the Nano Banana 2 AI image generator directly into its satellite and 3D imagery interface, letting users type a prompt and output a photorealistic, AI-modified version of any real location on Earth. Bellingcat founder Eliot Higgins posted a generated image showing a giant golden Trump statue placed over the White House within hours of launch. Google reversed the feature before the first full day of availability had passed.

The people who felt this fastest work in verification and conflict reporting

• Open-source intelligence analysts who use satellite imagery to confirm or deny military movements in active conflict zones
• Investigative journalists who cross-reference Google Earth visuals against on-the-ground photo and video evidence
• Educators who were the stated target audience and who now face students asking why the feature vanished

The rollback happened fast, but the feature existed long enough to produce shareable examples — and those examples are now permanently in circulation.

AI image generation and geospatial data were always going to collide, and this is only the first collision

Generative image tools are already embedded in consumer products used by hundreds of millions of people, and satellite imagery platforms are not far behind. The next attempt — from Google or a competitor — will come with better guardrails or better marketing, but the underlying capability is not going away.

What this tool was designed to do before it disappeared

• Generate AI-modified aerial views of real locations using text prompts
• Visualize completed real estate development projects over existing land
• Transform real urban environments into alternative or speculative scenarios
• Export photorealistic imagery grounded in authentic Google Earth data

Pricing was not established before the feature was retracted — check our directory for any future relaunch details.

The core limitation is the one that ended it: there is no technical barrier between legitimate use and fabricating evidence of events that never happened at real, identifiable locations.

Platforms like Midjourney and Adobe Firefly generate AI imagery but do not anchor outputs to verified real-world coordinates the way Google Earth does. That anchoring is exactly what made this feature more dangerous than generic image generation tools, and also what made it attractive to Google in the first place.

The line between geospatial verification and AI fabrication is now a product decision, not a technical one

Every week we track where AI tools are being deployed in ways that professional researchers, journalists, and analysts need to know about before they affect your work. We cover tools like this every Friday — subscribe here and we’ll send the best ones straight to you.