What's blocked automatically
Each photo and video is checked by Google Gemini, configured for our scenario. One call — multiple criteria at once:
1. Whether the photo actually shows a real person. Landscapes, cars, animals, screenshots, cartoons — rejected.
2. Whether it's not an AI-generated image. The model catches most cheap deepfakes. Not 100% (technology shifts), but economically makes faking significantly more expensive.
3. NSFW content. Nudity, sex content, graphic violence — blocked.
4. Masks and full face hiding. Not «bad» photos — they just don't serve the function of a primary profile photo.
Photos with friends are allowed. There is no requirement to be alone in the photo.
Personal verification
AI catches the obvious. Our team handles edge cases plus all reports from people. Response time hours to a day.
What we DON'T AI
No attractiveness-based matching. Algorithm doesn't score «how attractive» and doesn't show by «tier». Industry antipattern.
No video used for model training. AI runs once at upload for basic checks.
No AI to generate user content. No autofilled bios, no «AI will improve photos».
What about errors — false positives
AI verifications make mistakes. In that case — write to us at support (contacts on the «Contacts» page). A real person from our team reviews the content manually and, if AI was wrong, reverses the rejection. Standard response time — up to 24 hours.
Why this is structurally important
AI verification + mandatory video registration = a fundamental audience filter. Most fake profiles die at this stage. Passing is more expensive than the gain from a fake account.