- Google introduced Maps Imagery Grounding at Cloud Next, allowing users to composite AI-generated images into real Street View scenes via text prompt; the feature is US-only and in private preview.
- A new Aerial and Satellite Insights feature for Google Earth AI runs inside Google Cloud’s BigQuery and claims to reduce manual satellite image review from weeks to minutes.
- Two experimental Earth AI imagery models, available in Google Cloud’s Model Garden, are pre-trained to detect infrastructure objects — bridges, roads, and power lines — removing the need for custom model training.
- None of the three tools are in general availability; Maps Imagery Grounding is application-only, the imagery models are flagged experimental, and Aerial and Satellite Insights has not yet launched.
What Happened
Google unveiled three AI-powered imaging tools at its Cloud Next conference in April 2026, extending its Street View and Earth satellite data assets into AI-assisted creative and analytical workflows. The products — Maps Imagery Grounding, Aerial and Satellite Insights, and two Earth AI imagery models — were announced as limited previews with no general availability date given for any of them. Google positioned the releases at film production companies, urban planners, and infrastructure developers.
Why It Matters
The announcements add AI-generation and automated analysis layers on top of two of Google’s most extensive proprietary datasets: Street View, which covers more than 10 million miles of roads globally, and its satellite and aerial imagery archive. Geospatial AI has been an active area of development across the industry — Microsoft integrated aerial imagery into Azure Maps for developer use, and Palantir has built satellite-based monitoring workflows for defense and government customers — but Google’s move brings these capabilities directly into its existing BigQuery and Maps Platform developer ecosystem.
Technical Details
Maps Imagery Grounding is integrated into the Gemini Enterprise Agent Platform and composites AI-generated images with live Street View frames in response to natural-language prompts. Google’s stated demonstration example was a prompt reading: “generate an image of a futuristic spaceship hovering in front of the Washington Square Arch,” which the system resolves to a composited image grounded to the real location within seconds. The resulting still images can then be fed into Google’s Veo video model for animation.
Aerial and Satellite Insights operates directly inside BigQuery, meaning the imagery analysis pipeline does not require data movement outside Google Cloud. Google states that data analysts and urban planners currently review thousands of satellite images manually to detect landscape changes such as new construction; the tool claims to automate that detection. The two companion Earth AI imagery models in Google Cloud’s Model Garden are trained on specific infrastructure categories — bridges, roads, and power lines — and are designed to let developers skip the months-long process of collecting labeled training data and building custom classifiers.
Who’s Affected
Film studios and creative agencies are the primary stated audience for Maps Imagery Grounding, replacing or supplementing in-person location scouts that Google describes as expensive and time-consuming. Urban planners and data analysts working in Google Cloud environments are the target for Aerial and Satellite Insights, with construction site monitoring cited as a direct operational use case for infrastructure resource allocation. Software developers building geospatial products are the intended consumers of the Earth AI imagery models, which Google frames as a way to reduce time-to-deployment for object-detection applications.
What’s Next
Maps Imagery Grounding is accepting applications for its US-only private preview; non-US access and a broader rollout timeline were not disclosed. The two Earth AI imagery models are available now on an experimental basis in Google Cloud’s Model Garden. Aerial and Satellite Insights has not launched as of April 23, 2026, with Google indicating availability “in the coming weeks” and offering early-access sign-up for interested organizations.
