Google introduced an imaginative new feature to Google Earth.
Users could take real satellite imagery and use the company’s Nano Banana 2 image generator to transform the landscape through text prompts.
One day later, Google removed it.
The company said it had seen people sharing generated images that appeared to violate its policies and that the feature would remain unavailable while stronger safeguards were developed. �
TechCrunch
What Could the Feature Do?
The experimental tool allowed users to place AI-generated material over real geographic imagery.
Someone could theoretically ask the system to imagine:
A futuristic city in a desert
A restored historical monument
A different architectural plan
New green spaces
Conceptual flood defences
A fictional transport system
For architects, planners, educators and artists, the feature offered creative possibilities.
It could help people visualise alternative uses of land or communicate proposed developments more clearly.
But the same tool could also generate fabricated scenes that appeared to come from one of the internet’s most trusted sources of geographic evidence. �
TechCrunch
Why Critics Reacted So Quickly
Google Earth images are frequently used by:
Journalists
Researchers
Investigators
Humanitarian organisations
Students
Urban planners
Members of the public
People often treat a Google Earth screenshot as evidence that a place, building or physical event exists.
The new feature risked weakening that assumption.
A fabricated screenshot might appear to show:
A military installation
Damage from a natural disaster
Illegal construction
Environmental destruction
A refugee camp
A political demonstration
Buildings that never existed
Even where generated images were marked, screenshots could be cropped or reposted without the original context.
Critics described the feature as a potential source of “geospatial slop”—synthetic geographic material spreading without reliable provenance. �
TechCrunch
Why Google Reversed Course
Google said professionals had used the tool for legitimate purposes.
However, it also observed generated screenshots being shared in ways that appeared to breach company rules.
The company therefore rolled back the feature while working on stronger guardrails. �
TechCrunch
The speed of the reversal is notable.
Technology companies often release experimental AI features and wait for public harm to reveal the weaknesses.
In this case, the risks became obvious within approximately one day.
The Difference Between Creative Maps and Evidential Maps
There is nothing inherently wrong with fictional geography.
People create imagined cities, historical reconstructions and speculative maps for education and entertainment.
The danger appears when fictional imagery inherits the visual authority of an evidential platform.
A generated city displayed inside an obvious design application is understood as a concept.
The same city displayed within Google Earth may be interpreted as an actual satellite observation.
Context changes credibility.
The Bigger AI Lesson
Trusted platforms carry accumulated authority.
When AI-generation tools are inserted into those platforms, the synthetic output may borrow that authority.
This applies beyond maps.
A generated answer inside a search engine can feel more credible than the same words posted anonymously.
A synthetic image inside a news-style interface can feel more authentic than one inside an art application.
Companies therefore need to evaluate not only what a model can generate, but what users are likely to believe because of where the output appears.
Mary Chuks’ Perspective
Creativity and evidence should not share the same visual container without unmistakable boundaries.
A tool can support imagination while preserving trust if it includes:
Permanent visible labels
Embedded provenance data
Separate creative workspaces
Restrictions on realistic sensitive events
Original-image comparison
Clear export warnings
The mistake was not imagining alternative geography.
It was allowing imagination to look too much like observation.
Practical Takeaways
Do not treat screenshots as geographic proof.
Open the location directly in a trusted mapping service.
Compare imagery from multiple dates and providers.
Look for AI labels and provenance data.
Ask who first uploaded the screenshot.
Verify dramatic geographic claims through credible reporting.
Preserve the distinction between conceptual visualisation and satellite evidence.
Conclusion
Google Earth helps people believe they are looking at the world.
That trust is precisely why an embedded image generator required unusually strong protections.
Artificial intelligence can help us imagine what a place might become.
It must not make it impossible to know what is already there.
Original Source and Further Reading
Original reporting: Lucas Ropek, TechCrunch, “Google Nixes Its Earth AI Feature One Day After Launch, Amid Criticism It Would Spread Misinformation,” published 31 July 2026. �
TechCrunch