Crawler Analytics

What training, live answer and user-triggered visits mean

CiteTrack AI crawl log classifying each AI bot visit as training, live answer or user-triggered, with verified and spoofed flags
The Crawl Log labels every AI visit by intent, training, live answer or user-triggered, and flags spoofed bots.

Every AI bot visit in CiteTrack is labelled with an intent, the reason the bot fetched your page. There are three, and they answer very different questions. Knowing them is the difference between “a bot showed up” and “I am appearing in live AI answers right now.”

The three intents

Training

The bot is collecting content to train or refresh a model’s general knowledge. Examples: GPTBot (OpenAI), ClaudeBot (Anthropic), Google-Extended (Google) and Bytespider (ByteDance). A training visit means your content may inform future answers, but it is not tied to any one live question. It is the most indirect signal, good for confirming the big crawlers can reach you, but it does not prove you are being cited.

Live answer

The bot is fetching your page to build or refresh the index an assistant answers from, pulling sources in near real time. Examples: OAI-SearchBot (powers ChatGPT search) and PerplexityBot. This is the most valuable crawler signal, because a live-answer fetch is a leading indicator that your page is in the pool the engine draws answers from. CiteTrack counts these as AI answer impressions.

User-triggered

A real person prompted the assistant to fetch your page right now, they pasted your URL or asked the assistant about it. Examples: ChatGPT-User (OpenAI) and Claude-User (Anthropic). This is the strongest “a human is engaging with my page through AI” signal there is: someone just put your page in front of an assistant.

Why the split matters

A plain crawler log lumps these together as “a bot visited.” The intent breakdown turns that into something you can act on:

  • Lots of training, little live answer or user-triggered? The big models know you exist, but you are not surfacing in live answers yet. That is an AEO problem, not a crawling problem.
  • Live-answer fetches climbing? You are entering the answer index, pair this with citation tracking to confirm you are actually cited, not just fetched.
  • User-triggered visits appearing? People are sharing or asking about your pages in AI. Note which URLs, those are your strongest content.

Where to see it

Open CiteTrack AI → Crawlers. The intent mix is shown for recent activity and you can filter the visit log by intent. The AI answer impressions tile on the Dashboard is the live-answer count over the last 24 hours.

An AI answer impression is a live-answer fetch, proof the engine pulled your page while assembling an answer. It is not the same as a citation. A fetch means you were considered; a citation (tracked under Citations) means you were named.

← All articles