On Thursday I scrolled through a feed that suddenly felt off-kilter. A nuclear plant appeared over Iran in Google Earth, and a plane tore into One World Trade Center. By Friday the feature was gone.
I’m here to walk you through what happened, why it mattered, and what it means for the faint line between useful tools and disinformation factories. You’ll get what I think and what you should watch for next.
On Thursday, Google put a text-to-image tool inside Google Earth and people immediately abused it
That was the real-world moment: a feature meant to help people visualize places began producing fakes that looked like documentary evidence. The tool generated images that millions trust Google Earth to represent, and some users turned it into an attention machine—planting a nuclear plant in Iran and staging a plane strike on a New York landmark.
Google’s response was blunt: roll it back while stronger guardrails are added. The company emphasized the generated images were watermarked and didn’t appear in the public Google Earth experience, but a watermark is a thin barrier—easy to strip with other AI tools. The result was fast-moving screenshots that felt official because they were rendered inside a Google product.
Can Google Earth generate fake images?
Yes. For a brief window, Google Earth included an AI prompt feature that could synthesize imagery on demand. It didn’t insert the images into everyone’s map layers, but it made it trivial for any user to create realistic scenes and export screenshots that looked like authentic satellite or aerial views.
Tonight I typed just one sentence into Google Earth and put refugees near the Mexican border. Then I planted a nuclear plant in Iran. What on earth is Google doing? Check my latest post here: https://t.co/N0h6CVECqS pic.twitter.com/WSjdcEXWJx
— 𝚑𝚎𝚗𝚔 𝚟𝚊𝚗 𝚎𝚜𝚜 (@henkvaness) July 30, 2026
A viral screenshot showed a refugee mass at the U.S. border and people noticed quickly
Someone typed a single prompt and the result was retweeted and shared as if it were a live feed. That tweet from Henk van Ess and the commentary from public figures like Alex Hirsch turned the feature into a cultural lightning rod.
When you see a credible image inside a trusted framework, your mental guard goes down. That’s why even obvious safeguards—watermarks, restricted sharing—can fail. A fake image inside a respected product operates like a cheap stage prop: it borrows authority and sells a story faster than you can fact-check it.
Why did Google remove the feature?
The company said it detected misuse that violated its policies and that professionals had found helpful uses, but the potential for viral harm outweighed the benefits. In short: they couldn’t accept the risk of fabricated scenes circulating as apparent evidence.
On Friday Google pulled the tool and promised stronger guardrails
That action was the immediate, real-world consequence: the feature was gone within 24 hours.
Rolling back a public-facing AI feature is rare—but not unprecedented—when a platform realizes the social externalities are unmanageable. You should expect tech firms to test and retract features more often as regulators, journalists, and users push back. This episode also shows why companies that chase broad AI integration can trigger backlash beyond Silicon Valley. For some executives, AI growth supports stock narratives; for others, it shapes labor leverage. The stakes are wages and influence, not just convenience.
There are practical takeaways if you care about evidence and credibility. First, check provenance: where did the image come from, and who published it? Second, look for multiple independent confirmations—satellite feeds, official statements, civil-society witnesses. Third, treat AI-created imagery coming from a mainstream product with skepticism until you can verify it.
I want you to keep a simple habit: when a startling image arrives from a trusted brand, pause and ask how it could have been produced. That single question buys you time to avoid amplifying harm, because once a fake spreads it acts like an open pressure cooker—dangerous pressure builds fast and splatters everywhere.
Google, X (formerly Twitter), and public figures like Alex Hirsch are now part of the conversation about where AI should be allowed to touch our shared maps and memories. The company’s fast rollback feels like a small victory; the deeper problem—how to design tools that are useful without being weaponized—remains unsolved. What will you watch for next?