# How We Detect AI Agents on Our Website (And How We Let Them Hand Us Money)

> A complete technical blueprint for cooperative AI agent detection, DOM canaries, honeypot auto-recovery, and agentic commerce telemetry.

## The 4-Layer Architecture
1. **Client-Side Physics & Automation Fingerprinting (agentDetection.ts):** Zero-dependency 3KB detector measuring pointer bezier curves, touch pressure, and keystroke dwell times vs synthetic 0ms clicks and CDP automation globals (__playwright, cdc_).
2. **Cooperative DOM Canary Handshake:** Semantic visually hidden `<section id="valstorm-ai-protocol" class="sr-only">` informing LLMs about /.well-known/ai-agent.json, window.__valstormRegisterAgent(), and anti-spam verification rules.
3. **Honeypot Auto-Recovery:** Automatic rescue of benign agent submissions typing "N/A" or "None" into hidden security fields (hp_company_url) with CRM tagging as ai_browser_agent.
4. **Machine-Readable Discovery (/.well-known/ai-agent.json):** Standardized manifest detailing operator contacts, programmatic POST /v1/inbound/submit action schemas, and direct Google Calendar scheduling URLs.

URL: https://www.valstorm.com/playbooks/how-we-detect-ai-agents-and-let-them-hand-us-money