Docs/ concepts/ AI Traffic Types
conceptv1stableVerified 2026-08-01

AI Traffic Types

Understand how SeeLLM classifies observed crawler requests, retrieval requests, and detectable referrals.

How SeeLLM classifies requests

SeeLLM classifies server-side requests using signals such as user agents, request headers, network information where available, and request behavior. Classification describes what was observed; it does not prove how a model used the page.

Read the Evidence Model for the definitions and limits behind each result.

Human traffic

Browser requests that resemble normal user navigation. Signals can include a standard browser user agent, HTML and asset requests, and referrer headers.

AI retrieval

A user-directed or search-oriented request when a platform exposes a distinguishable signal, such as ChatGPT-User or Claude-User. A retrieval request shows that the platform fetched a page in a live-use context. It does not by itself prove that the answer displayed a citation or quote.

Crawler traffic

Requests from bots used for indexing, discovery, or training. Examples include GPTBot and ClaudeBot. A crawler visit proves an observed request to a URL, subject to classification confidence; it does not show that the page appeared in an answer.

AI referral

A human visit with a detectable referrer from an AI product. Because some clients strip referrer headers, referral counts are a measurable lower bound rather than a complete total.

Classification signals

SignalWhat it can supportImportant limit
User agentClaimed crawler or retrieval identityCan be spoofed
Network informationVendor ownership where published and verifiedNot available for every platform
HeadersRequest context and content preferencesNot proof of answer output
Request patternsSupporting behavioral classificationCan resemble other automation
ReferrerDetectable AI-origin human visitOften stripped by clients

Interpret the funnel carefully

A crawler visit, a user-directed retrieval, a sampled citation, and an AI referral are separate observations. Compare them at the same URL and over a stated observation window. Do not infer citation or referral growth from crawler activity alone.