Imagery search (Experimental)

Imagery search leverages a new, experimental multimodal (vision-language) Earth AI model to search across satellite and aerial basemap imagery in Google Earth. Unlike manual visual inspection, where users must painstakingly scan the map to find specific features, this tool automates discovery. Imagery search lets you find visual characteristics and patterns and localize them on the map instantly.

Imagery search is available to users on Google Earth Professional and Professional Advanced plans. Unlock imagery search and other advanced features by upgrading your plan.

Screenshot of imagery search in Google Earth

You can search for visual features (in the US only) within or inside a user-defined area of interest — such as a polygon, placemark, region (county, zipcode), or your current camera view — by selecting the Imagery search toggle or by asking the system to invoke it.

Example prompts include the following:

  • [Toggle selected] "Find empty lots in East LA" (or "within my selected polygon")
  • [Toggle selected] "Find algae blooms in rivers in Texas"
  • [Toggle selected] "Find large commercial parking lots with solar canopies in Boise, Idaho"
  • "Using imagery search, find large parking lots close to highway exits in San Francisco"
  • "Using imagery search, find highways with signs of distress in NY County"

Resulting behavior and limitations

Imagery search is designed to be a lead generator aiming to speed up searches for uncommon scenes or objects, rather than a tool for definitive inventory.

  • Relevance limit: The system returns a maximum of 30 results, selected based on the highest relevance to your query.
  • Best match logic: The model attempts to find the best visual matches for your query within the specified area. For example, if you search for an object in an area where it does not exist, the system may still return the closest visual approximations found in that region.
  • Recall and accuracy: As an experimental feature, imagery search is continuously evolving. You can expect accuracy and recall to improve over time, but results are not guaranteed to be authoritative. Users should visually verify all results.

As an experimental feature, imagery search is constantly evolving. Your feedback is critical to its development. If you encounter inaccurate results, blocked queries, or have suggestions for improvement, use the in-product feedback mechanism to rate your experience. Your input directly helps us refine the model's accuracy and usefulness.