AI transparency report

Every photo is analyzed by Google Gemini 2.5 Flash, a general-purpose multimodal model, accessed via Google's paid API. The analysis happens in a single request per photo: safety check, face detection, and scoring are decided together. There is no second model, no human review, and no re-rating of stored photos.

• Content safety: nudity or sexually explicit content ⇒ automatic rejection. If the model fails to answer, the upload is rejected (fail-closed).
• Face presence: a real human face must be clearly visible; drawings, AI-generated faces, animals and objects are rejected.
• The score (0–10): an overall facial-aesthetics judgment plus five sub-metrics (jawline, skin, symmetry, harmony, definition), each 0–10 with one decimal.

The prompt asks for objective, critical scoring — average faces should land around 4–6. We do not post-process or curve scores.

Facial-aesthetics models inherit biases from their training data: performance can vary across ethnicity, age, gender presentation, lighting and camera quality. Scores are not stable measurements — the same person can get different scores from different photos. A score is an automated opinion about one image. It is not a scientific or medical assessment, not a measure of character, health, or worth, and it must not be used for employment, education, insurance, credit, or any consequential decision about a person.

• Your photo (re-encoded copy, EXIF/GPS stripped) — private storage.
• A 256px thumbnail and a heavily blurred watermark variant.
• The score, sub-metrics, rank and timestamps in our database.
• Your username (18+ is self-attested via checkbox; no date of birth is stored).
• The photo itself is sent to Google's Gemini API once, at upload time, to be rated. See Google's privacy policy for their handling.

• Consent is explicit and versioned before your first upload.
• Withdrawing consent (Settings → Danger zone) permanently deletes all photos, scores and your profile — immediately.
• Deleting your account does the same, including cancelling billing.

FACERANK operates as a transparency-first AI system under Regulation (EU) 2024/1689 (AI Act): users are informed that AI generates every rating (Art. 50 transparency obligations), biometric processing happens only with explicit consent, no emotion inference or categorization beyond aesthetic scoring takes place, and this page documents the system's purpose, model, data flows and limitations. This product is entertainment; we explicitly disclaim consequential-use applications.