Before you scan

Photos you choose or capture stay on your device until you tap Scan this photo.
We do not require an account, email, or sign-in.
Scan history (thumbnails and report metadata) is saved locally in your browser - you can export or delete it anytime from the History page.

When you scan

Starting a scan sends your compressed photo to our server at /api/scan. From there:

AI analysis (default) - When configured, the image is sent to Google Gemini (model: gemini-3.6-flash) for multimodal vision analysis. Google processes the image according to their API terms.
We do not store your photo on our servers. The image is used for the analysis request and then discarded. Server logs record timing and success/failure - not image content.
The full-resolution photo returned in your report stays in your browser (IndexedDB). It is not uploaded again unless you scan again.

How long data is kept

On our servers: Scan images are not retained after analysis. Anonymous operational logs (duration, model used, success/failure) may be kept briefly for debugging.
On your device: History entries persist until you delete them or clear browser storage. Up to 100 recent scans are kept in local storage.
Feedback reports: If you report a wrong result, we log the scan ID, breeds, and your note - never the photo itself.

Who processes your data

Google Gemini - Vision analysis when AI mode is active. See Google Gemini API terms.
Vercel - Hosts the application and API endpoints.

No advertising trackers or third-party analytics pixels are used in the scan flow.

Demo mode vs AI analysis

Not every scan uses live AI vision. You may see one of these labels on your result:

AI vision analysis: Your photo was analyzed by Google Gemini. Results reflect visible features compared against our breed library.
Demo mode: You scanned a built-in sample photo. Results are pre-generated for demonstration - not real vision analysis.
Local estimate (no AI): AI was unavailable (no API key configured, or the service failed). A local placeholder estimator produces illustrative results - they are not based on what the dog actually looks like. Treat these as demo output only.

We always label which mode was used. Demo and local-estimate results should never be presented as real AI vision output.

What “visual match” means

The percentage shown on your report is a visual similarity estimate - how closely visible features in your photo resemble a known breed profile. It is not a DNA certainty score, pedigree proof, or veterinary diagnosis.

90%+ - Strong resemblance to a single breed in the photo
70–89% - Likely mix or partial match
Below 70% - Uncertain or heavily mixed appearance

Photo quality, angle, grooming, lighting, and mixed ancestry all affect the score. In demo or local-estimate mode, the percentage is illustrative and should not be relied upon.

Share links

When you copy or share a scan link, the URL encodes the report text in a compact format. When possible, a compressed JPEG thumbnail of your dog is also embedded so recipients can see the photo in the shared report.

Share links do not upload photos to our servers - the thumbnail travels in the URL itself.
Anyone with the link can view the embedded thumbnail and breed estimate. Do not share if you are uncomfortable with that.
If the photo is too large for a safe URL length, the link falls back to text-only (you will see a note when copying).
Detailed owner guidance is rebuilt on the recipient’s device from breed catalog data when available.

Your controls

Delete individual scans or clear all history from the History page.
Export your scan metadata as JSON from History.
Report incorrect results directly from any scan report - no photo is included in feedback.
Clear browser storage to remove all locally saved photos and history.

Data & privacy.