If you've noticed your web traffic dropping but couldn't figure out why, you might be looking at the wrong metrics entirely. Arrivl launched on Hacker News this week with a focused pitch: track how AI agents discover, read, and interact with your website—something Google Analytics and similar tools fundamentally cannot do.
The Blind Spot in Your Stack
Here's the dirty secret about most web analytics platforms: they rely on JavaScript snippets that fire in browsers. AI agents don't execute JavaScript. They make raw HTTP requests, parse your HTML server-side, and never touch your client-side tracking code. When Perplexity sends a crawler to research your product or ChatGPT's Browse with Bing hits your API endpoints, none of it registers in GA4, Mixpanel, or Amplitude. Arrivl captures this traffic at the server level, giving you visibility into an entire channel that's been invisible until now.
What Actually Gets Tracked
The platform fingerprints agents by their user-agent strings, IP ranges, and request patterns. Major AI companies publish official bot identifiers—Arrivl maintains a registry of these and currently tracks over 30 platforms including ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, DeepSeek, and dozens more. For agentic browsers like Claude for Chrome or ChatGPT Atlas that use real browser instances, the system falls back to behavioral signals to separate them from human visitors. The platform shows a live dashboard breaking down traffic by agent: their data shows ChatGPT leading with 1,204 events in the tracked period, followed by Claude at 834, Perplexity at 482, and Gemini at 321.
Attribution When AI Sends You Customers
This is where it gets interesting for product teams and marketers. Arrivl doesn't just track which agents are crawling your site—it attempts to connect those visits to human conversions. When an AI recommends your product in a response and a user clicks through, the platform labels that traffic source so you can see which AI tools actually convert for your business versus which ones are just training on your content or answering questions without sending referrals.
AI Readiness Scoring
Every page gets graded for how well AI agents parse it. The system runs checks across seven categories of signals that affect how agents read and index your content—things like structured data, rendering requirements, and content accessibility. Each weak score comes paired with a specific fix recommendation. The company claims most issues can be shipped in an afternoon.
Installation Takes Minutes
Three options for getting started: AI-assisted install, manual snippet placement, or sending a pre-written brief to your developer. Once live, you get real-time notifications when new agents hit your site, traffic spikes occur, or pages stop getting crawled by previously active bots.
Pricing and Availability
Arrivl is free during beta. The company is explicitly looking for early users who want data before they want a demo—sign up now and keep full access as features ship. When the product exits beta, pricing details haven't been announced yet.
Key Takeaways
- Traditional analytics tools miss AI traffic because agents don't execute client-side JavaScript
- Arrivl captures agent visits server-side using user-agent strings, IP ranges, and behavioral fingerprinting
- The platform tracks 30+ AI platforms including ChatGPT, Claude, Perplexity, Gemini, Grok, DeepSeek, and more
- Attribution features connect AI referrals to human conversions for ROI measurement
- Page-level readiness scoring identifies parsing issues with specific fix recommendations
- Currently free during beta with no announced pricing timeline
The Bottom Line
This is the analytics layer that should've existed two years ago. As AI agents increasingly become the intermediary between your content and your customers, flying blind to that traffic isn't optional anymore—it's a competitive disadvantage. Arrivl's beta offering gives you the visibility to actually understand this channel before everyone else figures out they need it too.