Photo checks for dating apps and social discovery
Dating apps can use PhotoCheck to screen profile photos as they're uploaded. One call returns whether the photo looks AI-generated, whether it holds exactly one clear face, and what its metadata declares. Flag generated faces for review, ask for a different photo when there's no face, and leave everything else alone.
The problem
Fake profiles have moved from stolen photos, which a reverse image search finds, to generated faces, which it can't. A generated face is unique, so nothing matches. The pixel texture and the framing still give it away, and so does the file when its metadata survives.
The checks that matter
| AI score | Catches generated faces that a reverse image search will never match. |
|---|---|
| Face count | A profile photo with no face, or a group shot, is a prompt to ask for a better one, not a fraud signal. |
| Declared AI | Avatar-style images from in-app AI tools often say so in their metadata. |
A decision rule
Built on POST /api/v1/analyze. Thresholds are starting points; tune them to what a mistake costs you.
const r = await check(photo); // POST /v1/analyze
if (r.summary.ai_generated === "declared") return "reject:ai_label";
if (r.faces.count === 0) return "ask:no_face";
if (r.ai.score >= 0.8) return "review:likely_ai";
return "accept";Questions
Should a high score auto-reject a profile?
We'd send it to review instead. Heavy filters and compression push some real photos up the scale, and a wrongly rejected real user is a lost user.
Is face data stored?
No. Images are processed in memory and discarded; only the call count is recorded against your key.