Note ML model only runs on Desktop sessions

how-it-works.mdx

@@ -5,7 +5,7 @@ sidebarTitle: 'How it works'
 
 The Roundtable Proof-of-Human API provides continuous bot and fraud prevention without adding friction for your users. We solve this by combining two independent signal streams:
 
-1. **Behavioral analytics** analyze user interaction patterns such as typing cadence, mouse velocity, scroll rhythm, and touch gestures. A machine learning model evaluates these signals together to detect automated agent behavior.
+1. **Behavioral analytics** analyze user interaction patterns such as typing cadence, mouse velocity, scroll rhythm, and touch gestures. A machine learning model evaluates these signals together to detect automated agent behavior. The ML model only runs on Desktop sessions (Windows, macOS, Linux).
 2. **Device & network intelligence** checks for virtual machines, VPNs, datacenter ASNs, headless browsers, and similar flags.
 
 These streams feed real-time models that generate a single risk score that continuously updates throughout the user session.

using-the-dashboard.mdx

@@ -14,7 +14,7 @@ Click on `Sessions` in the left-hand navigation bar to see a table of all record
 - **Status** – Live or Complete, depending on whether there was any user activity within the last 30 minutes
 
 Biometric checks
-- **Agent behavior** – An ML model detected patterns across biometric data consistent with automated behavior
+- **Agent behavior** – An ML model detected patterns across biometric data consistent with automated behavior. The ML model only runs on Desktop sessions (Windows, macOS, Linux).
 - **Programmatic typing** – The user enters text in an extremely consistent way, consistent with programmatic entry
 - **Teleporting mouse** – The user's cursor jumps large distances with no intermediate movement
 - **No corrections** – A large amount of text is entered without any backspaces or correction events