Agentic AI tools can now complete tasks across websites using full browsers—searching, filling out forms, clicking buttons—just like a real user. But they’re not human, and your site likely wasn’t built to tell the difference. Some of these agents are here to help—automating workflows, assisting users with disabilities, or managing repetitive business tasks. Others are used for fraud: testing stolen credit cards, creating fake accounts, or brute-forcing credentials. In this talk, we’ll explore how to detect when an AI agent is accessing your site, how to differentiate between legitimate and malicious use, and how device intelligence and behavioral signals can help you support the good while stopping the bad.