Back to stories
product launchGenerated by an AI editor from the reporting and web sources listed on this page.

Google Earth's new AI image generator can fabricate convincing fake satellite imagery in seconds

A free, globally available text-to-image tool built into Google Earth lets anyone drop invented buildings, disasters, and refugees onto real coordinates—and researchers warn it could poison open-source verification.

Published The total reporting and web sources attached to this story.How many attached sources came from wider web research rather than monitored news feeds.The AI editor’s assessment of how strongly the attached sources’ quality and agreement support this article.

What matters

  • Google activated its Gemini-based image model ("Nano Banana 2") inside Google Earth's web app on July 30, 2026, free and globally available with no waitlist.
  • The tool generates images grounded in real satellite, aerial, and 3D data from the exact coordinates on screen, making fakes harder to detect than standalone AI generators.
  • Investigator Henk van Ess demonstrated fabricated images of refugees near the Mexican border, a bomb crater in Gaza, and a nuclear plant in Iran—each at true coordinates.
  • The feature is currently limited to the desktop web version and is absent from iOS and Android apps.
  • Google has acknowledged the altered images but has not yet detailed any guardrails or restrictions.

Launch facts

Price:
Free
Availability:
Global, no waitlist
Platforms:
Google Earth web (desktop browser)
Regions:
Worldwide

What happened

On July 30, 2026, Google turned on "Nano Banana 2"—its Gemini image model—inside the web version of Google Earth. The feature is global, free, and carries no waitlist. Anyone can zoom to any coordinates, click "create image," and type a text prompt describing what they want to see. The model then generates an image built on real satellite, aerial, and 3D data from the exact location on screen.

Within a day, open-source investigator Henk van Ess of Digital Digging published examples showing how easily the tool could be abused. By typing single sentences, he generated images depicting "refugees near the Mexican border," a bomb crater near a hospital in Gaza, and a nuclear plant in Iran—each placed at true coordinates with matching lighting and camera angle. Google responded to Digital Digging's AI-altered images, though the full statement was truncated in available reporting.

Unlike standalone generators such as DALL-E or Midjourney, which produce images on a blank canvas, Google Earth's tool grounds every output in the actual map data beneath it. That means a fabricated hospital or crater inherits the visual credibility of real satellite imagery at that exact spot.

Why it matters

The core problem is not that AI can generate fake images—that capability already exists. The problem is the grounding. Because the output is tied to real coordinates and rendered with the correct angle, lighting, and terrain, the fakes are far harder to distinguish from genuine satellite evidence. For journalists, fact-checkers, and open-source investigators who rely on Google Earth to verify events, this tool could fundamentally undermine trust in geospatial imagery.

Van Ess's demonstrations are especially alarming because they target sensitive, high-stakes contexts: border migration, conflict zones, and geopolitical flashpoints. A fabricated image of a bomb crater near a Gaza hospital, sitting at the correct coordinates with realistic shadows, could spread rapidly on social media before anyone has time to debunk it.

The rollout also raises product-safety questions. Google made the tool available worldwide with no waitlist, no access restrictions, and no apparent safeguards tailored to sensitive locations. For now the feature is limited to the desktop web version and is absent from the iOS and Android apps, which somewhat limits casual reach—but the web version is freely accessible to anyone with a browser.

What to watch

  • Google's policy response. Google has acknowledged the altered images, but it remains unclear whether the company will add guardrails, watermarking, or location-based restrictions.
    -Adoption by bad actors. Watch for fabricated Google Earth screenshots circulating on social platforms during breaking-news events.
  • Verification countermeasures. Whether fact-checkers and platforms develop tools to detect AI-altered geospatial imagery.
  • Regulatory attention. The combination of real coordinates and AI-generated content may attract scrutiny from policymakers focused on disinformation.

What to do next

Developers

Experiment with the Google Earth web image tool to understand its output quality and identify potential detection signals (metadata, artifacts, coordinate mismatches).

Detection tooling will be needed if fabricated geospatial images proliferate.

Founders

Evaluate opportunities for verification and provenance tools that can authenticate satellite imagery and flag AI-generated alterations.

A new market gap is opening for geospatial deepfake detection.

PMs

Audit your product's reliance on Google Earth imagery for any user-facing features and consider adding provenance labels or disclaimers.

Users may encounter fabricated imagery that looks authoritative.

Investors

Track startups in geospatial verification, content provenance, and deepfake detection as demand is likely to grow.

The grounding of AI fakes in real coordinates creates a distinct threat vector that existing deepfake tools may not address.

Operators

Brief your communications and trust-and-safety teams on the risk of fabricated Google Earth screenshots during breaking-news events.

Rapid-response teams need to know that geospatial imagery can no longer be assumed authentic.

How to test

  1. 1Open Google Earth in a desktop web browser at earth.google.com.
  2. 2Navigate to any location using search or by zooming manually.
  3. 3Click the 'create image' button (or equivalent prompt-based UI element).
  4. 4Type a text prompt describing a scene (e.g., 'a large crater next to a building').
  5. 5Observe the generated image and note how it integrates with the real terrain, lighting, and angle.
  6. 6Try generating images at sensitive or well-known locations to assess whether any guardrails trigger.

Caveats

  • The feature is only available on the desktop web version, not on mobile apps.
  • Google may add or remove guardrails at any time, changing observed behavior.
  • Generating fake imagery of sensitive locations raises ethical concerns even for testing purposes.