Is Google Earth's "Create Image" Feature Gone in 2026? AI Satellite Tool Shutdown and Google's Response

If Google Earth's Create Image button suddenly vanished, you probably didn't do anything wrong. The AI tool launched on web Google Earth July 30, 2026, and was pulled in under 24 hours — it could generate highly realistic fictional scenes at real coordinates. Even labeled AI images could mislead once screenshotted and shared.

As of August 3, 2026, Create Image is no longer accessible, and Google has not set a relaunch date. Below: what you'll see in the interface, then what Google's public statements confirm.

1. Current Status: Create Image Is Gone

If you search for "Create Image" in Google Earth on the web — or look for the AI generation button you used before — you'll likely find it missing. Google announced the rollback on July 31, and the web entry was removed shortly after.

Important distinction: what disappeared is the user-facing AI entry — not Google Earth's official basemap being AI-replaced. Terrain, aerial imagery, and 3D buildings still come from existing sources. Create Image was a prompt-driven overlay, not a rewrite of underlying map data.

2. What Was It? Why People Called It an "AI Satellite Tool"

The official name is Create Image, powered by Nano Banana 2. Users entered text prompts at any location to generate or transform photorealistic scenes on satellite, aerial, or 3D imagery — tied to real coordinates.

People called it an "AI satellite tool" because output looked like satellite views at real places. That's user shorthand — not proof Google replaced global basemaps with AI. Generated images were an overlay, separate from official survey data.

3. Why It Was Shut Down: Fake Scenes + Screenshot Spread

Criticism mounted fast: the tool could produce misleading geographic imagery — fake disasters, conflicts, construction, and public events. OSINT researchers demonstrated policy-violating outputs within hours, and some sensitive prompts weren't blocked.

• Coordinate credibility → Scenes anchored to real locations, so viewers assume they're seeing on-the-ground footage.
• Low barrier to entry → A text prompt in the browser — no GIS expertise required.
• Screenshots leave context behind → The spread path doesn't pass through Google Earth's own interface.

In its rollback statement, Google said it had seen people sharing screenshots of generated imagery that appeared to violate platform policies. That's the crux of the risk: the problem isn't just "can it generate?" — it's how screenshots propagate on social media without watermarks or interface context.

4. How to Read Google's Official Response

Google published statements via X and related channels. The core points below are drawn from publicly available, principle-level language — not a full formal announcement:

Google had noted SynthID watermarks on generated images, verifiable via Gemini or Google Lens. At rollback, the company said stronger guardrails are needed before re-enabling the feature — with no relaunch timeline disclosed.

5. Why Watermarks Aren't Enough

Watermark and metadata verification only work when viewers still hold the original file and actively choose to check it. The typical spread path looks like this:

• User generates a scene inside Earth → screenshots it to X, Reddit, or group chats
• Screenshots carry no readable SynthID — ordinary readers can't verify in one click
• A caption naming the location further boosts visual credibility

So screenshot propagation is the hardest link to defend. In-app AI labels do little once content leaves the platform — a key reason Google chose fast rollback over watermark tweaks alone.

6. Where to Watch for a Relaunch

As of August 3, 2026, there is no confirmed relaunch date. Watch Google Earth release notes, the Google official blog and X accounts, Help Center documentation, and trusted tech media (TechCrunch, Engadget) for updates.

Still have questions?

Geospatial AI and Verification: A Practical Role for Mac mini

For researchers running local OSINT toolchains or image-verification pipelines, a stable, low-power node beats relying on a single cloud interface. Mac mini M4's unified memory and Neural Engine handle image comparison and metadata extraction on macOS; Gatekeeper, SIP, and FileVault isolate sensitive geospatial data.

At roughly 4W idle, Mac mini suits 24/7 imagery pulls. Python, Docker, and GDAL work out of the box. For provenance and AI verification on reliable hardware, Mac mini M4 is a cost-effective start — explore Mac mini cloud hosting and build your analysis environment on a node you control.