Google Earth AI image generator raises misinformation fears

Google Earth launched a new AI image generator globally on its web version on July 31, 2026, allowing users to create reality-warping images from satellite, aerial, and 3D imagery with simple text prompts. The feature, called Nano Banana 2, quickly raised concerns about its potential to generate convincing misinformation, as demonstrated by researcher Henk van Ess of Digital Digging, who produced images depicting refugees near the Mexican border and a bomb crater near a hospital in Gaza.

Van Ess's experiments highlight how the tool can fabricate scenes that appear to be authentic satellite views, potentially misleading viewers. Google responded to these demonstrations by emphasizing its commitment to combating misinformation, stating that every image created with Nano Banana in Google Earth includes a SynthID digital watermark. The company advises users to verify suspicious images by asking the Gemini app or using Lens in Search to check for AI generation, and also suggests using the "@verifyai" tag in Gemini.

Van Ess recommends additional verification steps, such as comparing images to other satellite platforms like Sentinel-2 or Landsat, and checking orbital data including the stated capture time and satellite identity. Despite these safeguards, the watermark may be missed or bypassed; Digital Digging managed to fool Hive's AI detector with an AI-altered video from Google Earth, underscoring the challenge of reliably identifying synthetic content.

Google's blog post pitches Nano Banana 2 as a tool for visualizing historical sites and real estate projects, but its capacity to generate realistic yet false imagery poses risks for spreading misinformation. The feature's global rollout on the web version of Google Earth marks a significant expansion of AI-powered image manipulation accessible to the public, raising questions about the integrity of geospatial information.

The incident with Digital Digging's images, shared on social media, illustrates the ease with which the tool can be used to create deceptive content. While some outputs, like a hybrid of the Sphinx and the Statue of Liberty, are clearly fake, others could be believable enough to deceive viewers, according to van Ess. The reliance on watermarks and AI detection tools may not be sufficient, as demonstrated by the failure of Hive's detector in one case.

Moving forward, the effectiveness of Google's verification ecosystem will be tested as users explore the boundaries of this technology. The combination of SynthID watermarks, Gemini checks, and manual verification against other satellite data provides a multi-layered approach, but the potential for misuse remains a concern.