Identifying Agentic Automation With Behavioral Telemetry: Part 2
Akamai Technologies, Thursday, August 27th, 2026
Akamai research uses behavioral embeddings and mouse telemetry to detect sparse AI agent traffic that bot detection misses.
Part two of Akamai's research applies behavioral embeddings and mouse-movement telemetry to identify autonomous AI agent traffic that conventional bot detection overlooks. The two-stage architecture combines request-level and session-level detection, achieving a 0.981 ROC-AUC and 92.4% agentic recall on request sequences containing at least five mouse events.
The team tested whether the models generalize across different customers rather than overfitting to a single site's traffic patterns. The post characterizes what agent activity actually looks like in the telemetry data, and closes by outlining where the research goes next.