Short answer

AI assistants send visitors with a referrer such as chatgpt.com or perplexity.ai, or with a utm_source value they add to links. Group these sources into one "AI" channel, then watch users, landing pages and engagement per assistant. PubPulse does this automatically in its AI Traffic report.

AI assistants answer questions directly, but when they cite a source, some readers click through. For many publishers this traffic is still small, but it grows month after month and it behaves differently from search traffic. You cannot manage what you do not measure, so the first step is to see it clearly.

How AI assistants send visitors

There are two signals:

  1. The referrer. When a reader clicks a link in ChatGPT, Perplexity, Claude, Gemini or Copilot, the browser usually sends the assistant's domain as the referrer: chatgpt.com, perplexity.ai, claude.ai, gemini.google.com, copilot.microsoft.com.
  2. UTM parameters. Some assistants add utm_source=chatgpt.com or similar to the links they show. This survives even when the referrer is stripped.

Some clicks arrive with neither, for example from desktop apps that open links without a referrer. Those land in "direct" in every analytics tool, so treat your AI numbers as a minimum.

Assistants worth tracking

AssistantTypical referrerNotes
ChatGPTchatgpt.comalso utm_source=chatgpt.com on many links
Perplexityperplexity.aicites several sources per answer
Claudeclaude.ailinks appear when web search is used
Google Geminigemini.google.comseparate from Google Search
Microsoft Copilotcopilot.microsoft.comseparate from Bing
DeepSeek, Mistral, Grok, Meta AItheir chat domainssmaller but growing

Setting it up in GA4

In GA4 you create a custom channel group with a regular expression on the session source that matches these domains, and place it above "Referral". It works, but it applies only from the day you create it and has to be maintained when new assistants appear.

Setting it up in PubPulse

Nothing to configure: every visit from these sources is classified with the medium "ai" and appears in AI Traffic → Overview, with users, sessions, pageviews and landing pages per assistant, and in the real-time view. You can combine it with any filter, for example AI traffic from Germany on mobile.

What to measure

  • Share of total traffic per week: is the channel growing?
  • Landing pages: which articles are cited. They tell you what assistants consider the best answer on your site.
  • Engagement: time on page and pages per session. AI visitors often arrive with a precise question; a high bounce rate can mean the page answers it too late.
  • Assistant mix: Perplexity and ChatGPT behave differently; track them separately.

How to be cited more often

  1. Answer first. Put a two-to-three-sentence answer at the top of the page, then the detail.
  2. Use structured data. FAQPage, HowTo, Article and Organization markup help machines understand the page.
  3. Let AI crawlers read you. Check that robots.txt does not block GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended if you want to appear in their answers.
  4. Publish an llms.txt. A short plain-text summary of your site with links to the key pages.
  5. Keep facts current. Assistants prefer pages with dates, numbers and named sources.

Then watch the AI Traffic report for the pages you changed. With PubPulse the effect is visible the same day, not after a reporting delay.

Try PubPulse free for 30 days

Every feature, no pageview limit, no credit card. Paste one line of code and your first visits show up within seconds.

Start your free trial

or click around the live demo first, no sign-up needed