How Do You Analyze AI Traffic to Your Website?

If your organic search traffic has been declining, part of the reason may be that AI is replacing search itself.

As ChatGPT, Gemini, and Perplexity become mainstream, users are asking AI before they ever open a search engine. By the time AI returns an answer, the user never visits your site.

The problem: none of this is recorded in GA4.

What’s Happening Outside the Browser

When AI generates a response, it crawls content across the web.
Your site is likely being read — without you knowing it.

But GA4 only measures users who arrive via browser.
AI crawler requests are never counted as sessions.
They don’t appear in your dashboard at all.

GA4 doesn’t see the requests hitting your server.

AI Crawlers Can Be Identified

AI services send requests using distinct User-Agent strings.

AI ServiceCrawler Name
OpenAIGPTBot
Anthropic (Claude)ClaudeBot
MetaMeta-ExternalAgent
ByteDanceBytespider
PerplexityPerplexityBot

These are retrievable from server logs or CDN-level access data (e.g., Cloudflare). They never appear in GA4.
But the data exists.

AI Traffic Has Different Roles

Not all AI crawler activity means the same thing.
EdgeShaping classifies AI access into four distinct types:

TypeShareMeaning
Training57.7%Collecting data to train AI models
General crawl28.7%Periodic indexing updates
Search / RAG11.5%Knowledge base and search-augmented retrieval
User-triggered2.1%Real-time reference in response to a user’s query
Based on actual data from mare-interno.com

User-triggered access is the smallest slice — but the most significant.
It represents the closest equivalent to traditional search intent.

If user-triggered crawls to a specific page are increasing,
that page is likely entering AI’s answer candidate pool.

2.1% of total AI traffic
this is the number that maps most directly to lost search visits.

Making It Visible with EdgeShaping

EdgeShaping is a measurement framework for structuring and analyzing AI crawler access to your site.

With it, you can answer:

  • Which AI services are crawling which pages, and how often?
  • Did crawl frequency change after a content update?
  • Can you separate human traffic KPIs from AI traffic KPIs?

There’s no direct way to measure whether AI is citing your content.
But you can measure whether AI is reading it.

No reads. No citations.
Building a “readable state” is the starting point of AI-era content strategy.

Expanding Measurement Beyond the Browser

Tool What It Covers
GA4Browser-based user behavior
Search ConsoleGoogle search impressions and clicks
EdgeShapingAll HTTP requests including AI crawlers

If search traffic is falling, you need to measure where it went.
The answer is outside the browser.

EdgeShaping is an AI measurement design framework proposed by mare interno LLC. It identifies and analyzes 65+ bot types including AI crawlers.