I build an agentic traffic tracker that uncovered 58% more dark traffic than standard traffic
I have been seeing weird patterns in our data lately. Direct traffic is spiking, but no clear source attribution. Users are landing on deep pages, converting fast, and mentioning they found us through ChatGPT or Perplexity, but GA4 just shows everything as direct.
This pushed me to build a tracking solution that identifies AI-origin behavior patterns, prompt-shaped visits, and crawl signals. I tested it across several sites and consistently found 58% more dark traffic than I was catching before. Most came from AI assistants with no proper referrer data.
The key was setting up server-side detection for agent signatures, analyzing session patterns that match AI-driven discovery, and creating attribution models for non-traditional referrers. I did this with limyai.