🌍

PRIV-LOC

Geolocation Privacy Challenge

Challenge yourself against state-of-the-art AI models in a real-time geolocation game. Can you identify photo locations better than Claude, GPT, Gemini, and Qwen?

5000
Max Points
8
AI Models
∞
Challenges
🎯

How to Play

Click on the map where you think the image was taken, then click Select to confirm your guess. The closer you are, the higher your score!

📍 Click to guess location
⚡ Real-time scoring
🏆 Beat the AIs
🤖

Your Competition

Face off against cutting-edge AI models from leading tech companies:

Claude Anthropic
GPT OpenAI
Gemini Google
Qwen Alibaba
📊

Scoring System

Earn up to 5000 points per image based on accuracy. Distance matters!

5000
Perfect guess
3000+
Within 100km
1000+
Within 1000km
Score = 5000 × e−distance⁄1492.7

🏆 Ready for the Challenge?

This isn't just a game—it's cutting-edge research in AI privacy and geolocation. Your performance helps us understand how humans compare to AI in visual location detection.

🔬 Contributing to AI research
🔒 Privacy implications study
🌐 Human vs AI comparison

📚 Cite This Research

If you use this work in your research, please cite:

@inproceedings{grainge2025geolocation,
  title     = {Assessing the Geolocation Capabilities, Limitations and Societal Risks of Generative Vision-Language Models},
  author    = {Grainge, O. and Waheed, S. and Stilgoe, J. and Milford, M. and Ehsan, S.},
  booktitle = {Proceedings of the AAAI 2025 Fall Symposium on AI Trustworthiness and Risk Assessment for Challenging Contexts (ATRACC)},
  year      = {2025},
  publisher = {AAAI Press},
  address   = {Arlington, Virginia, USA},
  note      = {Forthcoming}
}