# Log AI Requests and Referrals at the Edge

> Canonical source: [Log AI Requests and Referrals at the Edge](https://www.answer.cloud/pages/how-to-log-ai-bot-and-llm-referral-traffic-at-the-edge-vendor-agnostic-hosted/)

Collect request evidence with a working integration and keep classification and referral limits explicit.

Server or edge analytics can record requests that do not execute a browser tag. Browser analytics remains useful for human sessions and subsequent interactions. Neither source observes every AI-assisted event.

## Define the event

Record the request time, URL, method, status, observed user agent, referrer when supplied, and the fields required by your analytics destination. Handle personal data according to your actual collection and retention process. User-agent classifications can be spoofed, and referrers can be absent.

## Use a complete integration

For answer.cloud on Cloudflare, follow the [Worker guide](https://www.answer.cloud/docs/cloudflare-worker-agent-analytics-hosted/). It forwards the origin response and schedules the analytics POST with background work. A code sample containing only a “send to log store” comment does not record an event.

Other providers require their own supported log or edge integration. Verify that events arrive and that normal website responses are preserved.

## Interpret the data

A crawler request is not a citation. A referral is a visit, not proof that every answer was observed. Keep request categories, human sessions, and controlled citation tests separate. See [visibility metrics](https://www.answer.cloud/pages/llm-visibility-metrics-the-kpi-playbook-for-seo-and-marketing-leaders-hosted/).