Circle to Search

Why It Exists

Traditional search requires typing. Circle to Search addresses the need to understand visual context -- recognizing that users often want to know about something they see, not something they can describe with text.

How It Works

Circle to Search uses the on-device Gemini Nano vision model to identify and classify objects in the screen capture. The model processes the visual input entirely on the device to preserve privacy. The identified object is then used as a search query, which is sent to Google's search infrastructure for results. On HyperOS AI, the feature uses a similar on-device vision model with cloud-assisted search enhancement.

Everyday Use Cases

• Search for a landmark you see while traveling
• Identify a product in a photo to find reviews
• Circle text in a foreign language to translate it instantly
• Find the source of an image you see online

User Workflow

• Open any app where you see something you want to search for
• Touch and hold the Home button or swipe up from the bottom
• Circle, highlight, or tap the object, text, or image you want to search
• Google displays search results for your selection

AI Processing Flow

Uses on-device Gemini Nano vision model for object recognition, then sends query to Google Search.

Inputs / Outputs

Inputs:

• Visual selection
• Screenshot

Outputs:

• Search results cards
• Links
• Fact cards

Known Limitations

• Only available in English and 11 other languages at launch

Unsupported Scenarios

• Does not work in apps with DRM-protected content
• Cannot search for text in low-light images

Performance Notes

Processing typically completes in 2-5 seconds on supported devices with NPU acceleration.

Available On

Shows where this feature is available and how its AI processing works on each platform. Availability may vary by device.

Research Status

Confidence and verification reflect how complete documentation is. Fields may show Not assessed or Not yet verified while research is ongoing - this flags gaps, not product deficiencies.

Research Notes

Caveat: Google states the feature requires Google account sign-in for personalized results. Source: Google I/O 2024.

Details

Capability Mapping

This feature maps to canonical capability family: Visual Understanding.

Canonical capability: Visual Understanding.

• Visual Understanding - Interprets what is in view and provides contextual understanding or actions.

Grounded in 2 device(s) on Gemini Nano, Xiaomi HyperAI + Google AI, HyperOS AI.