Here’s the problem with putting an AI image generator in Google Earth
TL;DR
A single text prompt is now enough to generate reality-warping scenes from Google Earth's satellite, aerial, and 3D imagery. Digital Digging's Henk van Ess produced images showing supposed refugees near the Mexican border and a bomb crater beside a hospital in Gaza. Google points to SynthID, the digital watermark embedded in every Nano Banana image, which the Gemini app or Lens in Search can detect. Whether that check happens in a fast-moving social feed is the open question.
Nauti's Take
For research, urban planning, and storytelling, a prompt-driven image generator running on real geospatial data is genuine progress. The risk arrives in the same step: if geographic reality is this easy to rewrite, disinformation gets a very convincing template.
SynthID only helps when someone actually checks, and that rarely happens in a fast feed. Anyone publishing visual evidence should budget for provenance checks from now on.
Sources