To track referral traffic from ChatGPT, Perplexity, Gemini, and other AI assistants, you identify the referrer hostnames those tools send visitors from, build a custom channel or segment for them in your analytics, and reconcile that against your server logs — because a large share of AI-driven visits arrive with no referrer at all and get misfiled as direct traffic. Done right, this turns an invisible source of visitors into a measurable channel you can actually report on.

Why AI traffic hides in your analytics

When someone clicks a citation inside an AI answer, their browser may or may not pass a referrer. Assistants that live in native apps, that strip the referrer for privacy, or that open links in an in-app browser frequently send visitors with an empty referrer field. Your analytics has nowhere to file them, so they land in Direct alongside people who typed your URL by hand. The result is a real and growing channel that looks like noise. This is the same visibility gap covered in measuring AI traffic in your analytics, viewed here specifically through the lens of attribution.

Step one: know the referrer hostnames

When AI tools do pass a referrer, they use recognizable hostnames. The ones worth watching in 2026 include:

Treat this as a living list rather than a fixed one — assistants change domains and add surfaces over time, so revisit it periodically. Google AI Overviews are a special case: those clicks usually appear as ordinary Google organic traffic, not as a distinct referrer, which is why appearing there is tracked differently from a referral. If Overviews are your goal, focus on appearing in Google AI Overviews rather than expecting a clean referral signal.

Step two: build an AI channel or segment

In your analytics platform, create a grouping that captures those referrer hostnames as a single channel — call it AI Assistants or similar. Most tools let you define a custom channel group or a segment based on the referring domain matching a list of patterns. Once it exists, you can see AI-driven sessions next to organic, direct, and social, and watch the trend month over month.

Add a second layer with landing-page analysis: which of your pages actually receive AI referrals? The pages that earn citations tell you what kind of content assistants find quotable, which feeds directly back into how you write. Understanding that is the whole point of studying how ChatGPT chooses which websites to cite.

Step three: reconcile with server logs

Because so much AI traffic arrives referrer-less, analytics alone undercounts it. Your server logs tell a more complete story in two ways. First, the bot crawls — GPTBot, PerplexityBot, ClaudeBot, and others fetching your pages — show which assistants are reading your content in the first place, a prerequisite for ever being cited. Second, spikes in referrer-less human sessions to a specific page often correlate with that page being surfaced in an AI answer. Cross-referencing the two sources gives you a far truer picture than either alone. Server-side measurement is the reliable half of the equation, as detailed in server log analysis for AI bot crawling.

Step four: watch mentions, not just clicks

Attribution is not only about traffic. AI assistants often answer a user's question completely without sending a click at all — your brand can be recommended thousands of times and generate very few referrals. That means clicks understate your true AI visibility. To see the fuller picture, pair referral tracking with checking whether assistants actually name your brand in their answers, the practice described in monitoring whether AI assistants mention your brand. A page that gets cited often but clicked rarely is still doing its job.

Putting it together

A workable AI attribution setup has four moving parts: a maintained list of assistant referrer hostnames, a dedicated analytics channel built on that list, regular server-log reconciliation to catch referrer-less visits and bot crawls, and brand-mention monitoring to capture the visibility that never converts to a click. None of these is perfect on its own, but together they turn a blind spot into a reportable channel.

Before you can attribute AI traffic, assistants have to be able to read and cite your pages at all. Run the CheckMy.site scanner to confirm AI crawlers can access your content — then start measuring the visits they send back.