All shows / The Quiet Tell

The Quiet Tell cover art

The Quiet Tell

The Cars That Never Moved

A viral "destroyed radar" satellite image, a Google Earth button that lasted one day, and why a fake satellite picture is almost always a real one in disguise.

Ines and Jonah · 22 min · 23 Oct 2026

FactUnion VerifiedEvery fact on the episode's fact sheet was checked against its sources. Wrong facts were corrected. What this meansFacts checked 22 Oct 2026

Chapters (8)

About this episode

In early 2026, during the US-Israeli war with Iran, Tehran Times posted a "before vs after" satellite image on X that it said showed US radar equipment in Qatar completely destroyed. Fact-checkers at Misbar (Al-Araby TV) and DW Fact Check matched the "before" frame to a real Google Earth capture of a US naval base in Bahrain. One of the giveaways was a row of cars that hadn't moved. Then on July 30, 2026, Google put a "create image" button into Google Earth on the web. Researchers including Henk van Ess showed what it would make, and Google rolled it back the next day.

Ines and Jonah retell the work of the fact-checkers, analysts and reporters who caught these fakes, and say clearly whose work it was. They cover the twist in Manama, what an invisible watermark can and can't tell you, why a careless shadow check can make a real image look fake, and the habit that ties it all together: find the parent image. All the investigative work in this episode belongs to the people and outlets credited. The episode closes with an open satellite archive you can browse yourself.

Hosted by Ines and Jonah, performed by synthetic voices. Supported by STELQ.

Facts and sources

Every fact used in this episode, with the sources it was checked against. Download as data

198 of 198 recorded lines matched the checked script.

  1. AFP reported that Tehran Times, a state-aligned English daily, posted on X a ‘before vs after’ satellite image it said showed ‘completely destroyed’ US radar equipment at a base in Qatar.

  2. AFP reported researchers found the image was ‘an AI-manipulated version of a Google Earth image from last year of a US base in Bahrain,’ and that it ‘garnered millions of views as it spread across social media in multiple languages.’

  3. AFP cited ‘a row of cars parked in identical positions in both the authentic satellite photo and the manipulated image’ as one of the subtle visual giveaways.

  4. DW Fact Check said the ‘before’ image matches a genuine Google Earth capture dated February 10, 2025, of a US naval base in Manama, Bahrain, that the ‘after’ image shows building structures changing shape, and that the X post had been viewed over 950,000 times.

    Reworded to match its sources. Source corrected: the page first cited was read and is about something else; the sources listed state this.

  5. Misbar, in a fact-check dated March 1, 2026, said its team at Al-Araby TV identified the left image as a 2025 Google Earth screenshot of the Al-Juffair area in Manama, Bahrain, citing vehicle types and positions, a white vehicle beneath the dome and matching shadows.

    Kept on a lower-weight source (misbar.com); nothing we found says otherwise.

Show all 23 factsShow fewer facts

  1. The New York Times reported that video it verified showed an Iranian one-way attack drone striking a radome at the U.S. Navy’s Fifth Fleet headquarters in Manama, Bahrain.

    Reworded to match its sources. The original's satellite imagery claim and extra destroyed radome are unsupported. The evidence supports only the NYT-verified video of an Iranian drone hitting one radome.

  2. AFP reported that Tal Hagin, an information warfare analyst, flagged an AI-generated image purporting to show Israeli-US jets hitting a painted aircraft silhouette in Iran, betrayed by gibberish coordinates, and that AFP also detected a SynthID watermark in it.

  3. Rumor Scanner said its team used Google’s SynthID detection technology on the Iran ‘painted fighter jets’ image and confirmed it was generated with the assistance of Google’s AI.

    Kept on a lower-weight source (rumorscanner.com); nothing we found says otherwise.

  4. AFP reported that satellite company Vantor, via its spokesperson Tomi Maxted, said its own imagery helped confirm viral photos of Niamey airport’s civilian terminal on fire were fake, ‘almost certainly generated using AI.’

  5. Google’s July 30, 2026 launch post said: ‘Just zoom in to a place in Google Earth on web, tap “create image,” and type whatever you want to see,’ and that the feature was ‘available globally today for Google Earth web users.’

  6. The launch post said users could ‘generate custom images using Google Earth’s satellite, aerial, and 3D imagery alongside Nano Banana,’ with examples including Pompeii in 78 A.D., a Tokyo empty lot, a lakefront cabin and a sci-fi Mountain View campus.

  7. Google’s 7/31 update said: ‘we’re rolling back this feature in Google Earth while we work on implementing stronger guardrails,’ noting generated images ‘didn’t appear in the main Google Earth experience for others to see and were watermarked as AI generated.’

  8. Unite.AI reported the feature shipped July 30, 2026 and was removed July 31, about a day after being switched on for every browser user, and named Google Earth product manager Bryan Horowitz as the launch post’s author.

    Kept on a lower-weight source (techy101.com); nothing we found says otherwise.

  9. Henk van Ess, who publishes the Digital Digging newsletter, said of his tests: ‘I tried refugees at the Mexican border, a nuclear plant in Iran, a crash in Amsterdam, a hospital with a bomb crater in Gaza. Nothing was refused.’

    Kept on a lower-weight source (opensourceforu.com); nothing we found says otherwise.

  10. Unite.AI reported that van Ess posted a walkthrough hours after launch and that NPR’s own tests produced Iran’s Kharg Island on fire and a flooded US Capitol complex.

    Kept on a lower-weight source (unite.ai); nothing we found says otherwise.

  11. Bellingcat senior researcher Jake Godin told NPR: ‘It’s just streamlining the process, which I think is going to make it proliferate more.’

  12. Ross Burley of the Centre for Information Resilience said ‘Trust in satellite imagery has taken decades to build and could be irrevocably damaged overnight,’ and Evan Hill of the Washington Post called the opportunities for abuse ‘literally boundless.’

    Kept on a lower-weight source (forkast.news); nothing we found says otherwise.

  13. Forkast reported users circulated Create Image fakes of a Paris explosion, an Iranian nuclear site, a Russian bomb crater, an ISIS training ground in Syria and a flooded US Capitol.

    Kept on a lower-weight source (forkast.news); nothing we found says otherwise.

  14. Unite.AI described Google’s safeguards as SynthID, an invisible marker readable by Google’s detection tools that survives a screenshot but reports only when someone runs a check, plus prompt-level refusal of harmful topics.

    Kept on a lower-weight source (unite.ai); nothing we found says otherwise.

  15. SynthID embeds an imperceptible watermark in pixel values at generation, and a trained detector returns a confidence verdict of Watermarked, Uncertain or Not watermarked; a ‘not watermarked’ result does not prove an image is human-made.

  16. Google says its SynthID Detector portal, launched in 2025 and first tested with journalists and media professionals, is now expanding access to everyone to check content made with AI from Google or partners such as OpenAI, NVIDIA and Kakao.

  17. In tests on RapidEye orthoimages, using only the sun’s direction gave shadow-direction errors of up to 21.3° in high-oblique images, while adding sensor viewing geometry cut the error to under 4.0°.

  18. Vantor’s open-data satellite imagery is published under CC BY-NC 4.0 as cloud-optimized GeoTIFFs in a STAC catalog, a public archive anyone can browse.

    Kept on a lower-weight source (registry.opendata.aws); nothing we found says otherwise.

Removed during checking (2)

  • Niamey's Diori Hamani International Airport was attacked in January 2026, an attack that killed 20, and again on 18 June 2026 in an attack claimed by JNIM.

    Why it was removed: Sources only partly back this: The New York Times.

  • A June 2026 paper by WITNESS's Jacobo Castellanos Rivadeneira describes C2PA as 'emerging as the foundational infrastructure for verifiable provenance of digital content' while warning its signing chains carry surveillance risks without governance safeguards.

    Why it was removed: We could not find a public record about this either way.

More from The Quiet Tell

  1. The Ship That Pointed at the Pier

    Interference scrambled a grain ship's position across the Black Sea. Bellingcat found it anyway, using the one number the satellites couldn't touch: the direction its bow was pointing.

    13 Oct 2026 · 23 min Facts and sources

If you like The Quiet Tell, you might also like