Google has pulled a newly launched AI image-generation feature from Google Earth less than 24 hours after its debut, after journalists and open-source investigators warned it could make fabricating fake satellite evidence as easy as typing a prompt.

The abrupt rollback highlights growing concerns that generative AI is beginning to erode trust in one of the internet's most reliable forms of visual verification — satellite imagery used to investigate wars, natural disasters and geopolitical events.

AI-generated satellite "evidence" sparks immediate backlash

The feature, launched on July 30, allowed users of Google Earth's web version to zoom to virtually any location, click "Create Image," and generate entirely new satellite-style scenes using Google's Nano Banana AI model.

Within hours, researchers demonstrated how easily the tool could fabricate events that never happened.

Among the AI-generated images were a massive blast crater in Los Angeles, protesters gathered outside Google's Mountain View headquarters, a flooded U.S. Capitol and Iran's Kharg Island engulfed in flames — scenes that, if authentic, would represent major global news events.

Open-source investigator Henk van Ess also told NPR that prompts involving refugees along the U.S.-Mexico border and an Iranian nuclear facility generated images without being rejected, raising fresh concerns over how the technology could be exploited to spread convincing disinformation.

For journalists and open-source intelligence (OSINT) investigators, the implications were immediate. Satellite imagery has long been regarded as one of the most trusted sources for verifying military operations, humanitarian crises and war crimes because it has traditionally been difficult to manipulate. Researchers warned that AI-generated imagery could now blur the line between authentic evidence and sophisticated fabrication.

Google retreats as trust concerns outweigh innovation

Facing mounting criticism, Google removed the feature on July 31.

In a statement posted on X, the company said people "uniquely trust Google Earth for a reliable view of the world" and acknowledged that while geospatial professionals had found legitimate uses for the technology, some generated images appeared to violate company policies.

Google said it is withdrawing the feature while developing stronger safeguards but did not provide a timeline for its return.

The company noted that every AI-generated image carries its SynthID watermark and that the images were not integrated into the core Google Earth experience. Critics argued those measures offered little reassurance, warning that misleading images often spread widely across social media before anyone verifies whether they were AI-generated.

A new challenge for the AI era

The incident underscores how quickly generative AI is expanding beyond chatbots and artwork into tools capable of fabricating what has historically been considered objective visual evidence.

For years, satellite imagery has played a crucial role in exposing military movements, documenting humanitarian disasters and verifying breaking news. The swift backlash against Google's experiment suggests that, without stronger safeguards, AI-generated geospatial imagery could become a powerful new weapon for misinformation—forcing governments, journalists and investigators to rethink how digital evidence is authenticated.