AI Agent Analytics

The blind spot

AI agents like ChatGPT, Claude, and Perplexity browse the web on behalf of their users. They fetch pages, read content, and follow links — but many don't execute third-party scripts, and even those that do may not trigger analytics reliably. That means this growing category of traffic is often invisible to site owners.

Measure solves this with three complementary approaches: automatic detection from User-Agent strings, a simple self-reporting API, and discovery mechanisms that let agents find the reporting endpoint without any prior knowledge of the site.

How it works

Collect API

Send a POST request to report a page visit:

POST /api/collect
Content-Type: application/json

{
  "site_id": "abc123def456",
  "path": "/products/widget",
  "agent": "MyAgent/1.0",
  "referrer": "search-query",
  "utm_source": "ai-agent",
  "utm_medium": "referral",
  "utm_campaign": "product-research"
}

The endpoint returns a 1x1 transparent GIF with a 200 status. A 404 means the site_id is invalid. A 429 means the site has exceeded its plan limits.

Example: curl

curl -X POST https://lets.measure.events/api/collect \
  -H "Content-Type: application/json" \
  -d '{
    "site_id": "YOUR_TRACKING_ID",
    "path": "/page-you-visited",
    "agent": "MyAgent/1.0"
  }'

Discovery: .well-known/measure.json

Sites using Measure can host a /.well-known/measure.json file that tells agents how to report their visits. Agents should check for this file when visiting a new domain, similar to how they check robots.txt.

GET /.well-known/measure.json

{
  "analytics_provider": "Measure",
  "version": "1.0",
  "collect_endpoint": "https://lets.measure.events/api/collect",
  "method": "POST",
  "content_type": "application/json",
  "payload": {
    "site_id": "your tracking ID (required)",
    "path": "page path visited (required)",
    "agent": "your agent name (recommended)",
    "referrer": "referring URL or agent name (optional)"
  },
  "privacy": "https://example.com/privacy",
  "documentation": "https://lets.measure.events/docs/ai-agents"
}

The collect_endpoint is the full URL to POST visit data to. The payload object describes the expected fields and their purpose.

Discovery: X-Analytics-Ping header

Measure includes an X-Analytics-Ping response header on script and collect endpoint responses. The header contains the full collect URL with the site's tracking ID pre-filled:

X-Analytics-Ping: https://lets.measure.events/api/collect?site_id=abc123def456

When an agent encounters this header on any HTTP response, it can send a POST to the URL with a JSON body containing path and agent fields to report its visit.

For site owners

If you use Measure, AI agent traffic from known agents is detected automatically. No extra setup required — just add the tracking script as usual and agent visits will appear in your dashboard.

To support cooperative self-reporting from agents, you can optionally add a /.well-known/measure.json file to your site. Create the file at the root of your domain with the following content:

{
  "analytics_provider": "Measure",
  "version": "1.0",
  "collect_endpoint": "https://lets.measure.events/api/collect",
  "payload": {
    "site_id": "YOUR_TRACKING_ID",
    "path": "page path visited",
    "agent": "agent name"
  }
}

Replace YOUR_TRACKING_ID with your site's tracking ID from the Measure dashboard.

Privacy guarantees

Agent analytics follow the same privacy-first architecture as all of Measure. We store aggregated counters only — never individual requests, IP addresses, or session identifiers.

Automatically detected agents

Measure currently recognizes the following AI agents by their User-Agent strings:

Agents not on this list can use the agent field in the collect API to self-identify.

Building an AI agent?

Help site owners understand their AI traffic. It takes one HTTP request.

Questions? Reach out at support@measure.events.