Skip to content

Articles / Proving value

How do you track visits from AI answers? What analytics can and can't see

Google Analytics 4 counts AI assistant referrals in their own channel. App clicks, AI Overviews and unclicked answers stay hidden, so the count is a floor.

Published
Reading time
8 min read
By
AmpliRank
Tags
AI Referral Visits · Google Analytics · Reporting

Track visits from AI answers in Google Analytics 4’s AI Assistant channel, which groups visits whose referrer is an AI assistant, and in a custom channel you define with the assistants’ domains. Treat the count as a floor: clicks that arrive without a referrer land in Direct, clicks from Google’s AI Overviews and AI Mode are counted with Google search, and an answer that names you without a click leaves no trace at all.

In short

  • Google Analytics 4’s default channel group has an AI Assistant channel for visits referred by AI assistants. It excludes Google’s AI Overviews and AI Mode.
  • A visit with no referrer and no campaign tag is counted as Direct, whatever sent it.
  • Clicks from AI Overviews and AI Mode come from Google Search, and Search Console reports them together with ordinary web search.
  • A custom channel group with your own list of assistant domains gives you a definition you can see and keep stable, and it applies to past data.
  • Report AI referral visits as “at least this many”, next to how often engines name you, and call any change observed, not caused.

Why does this come up?

It comes up because leadership asks how much traffic AI sends, and the honest answer has two parts: the visits you can count, and the influence you can’t. An AI referral visit is a visit that arrives from a link in an AI assistant’s answer, identified by its referrer or a tag. Most of what an answer does to a buyer’s shortlist happens before any click, or without one.

The guide to reporting AI visibility to leadership puts referral visits in the brief as supporting evidence, under the visibility rate. Can AI visibility be tied to pipeline? covers what those visits can and can’t prove about deals. This article is the setup: where the visits show up, how to count them consistently, and where they hide.

Where do AI visits show up in Google Analytics?

AI visits show up in the AI Assistant channel when the browser passes an AI assistant’s address as the referrer. Google Analytics Help’s default channel group page says the medium is set to “ai-assistant” when the referrer matches a list of AI assistants, and that the channel excludes Google’s AI Overviews and AI Mode (checked 14 November 2026). The page doesn’t publish the list itself.

Every other kind of AI-influenced visit lands somewhere else. This table follows each one to where Google Analytics 4 counts it:

What the buyer did What reaches your site Where Google Analytics 4 counts it
Clicked a link in an AI assistant’s answer, in a browser A referrer with the assistant’s domain AI Assistant, if the domain is on Google’s list; otherwise Referral
Clicked a link that opened without a referrer, for example from an app No referrer, no tag Direct
Clicked a source link in AI Overviews or AI Mode A click from Google Search Organic Search, mixed with ordinary search clicks
Read your name in an answer, then searched for you A search click, often on your brand name Organic Search
Read your name, then typed your address No referrer Direct
Read your name and never visited Nothing Nowhere

Only the first row is an AI referral visit you can count. The rest is influence that analytics either files under another channel or never sees.

Why do some AI clicks arrive as Direct?

Some AI clicks arrive as Direct because they reach your site without a referrer. The referring page decides what the browser sends: under the no-referrer policy, MDN’s Referrer-Policy reference says requests “do not include any referrer information”. Google Analytics counts a visit whose source is “(direct)” and whose medium is “(not set)” or “(none)” as Direct.

Even when a referrer arrives, it says little. Under the browser default, a cross-site link sends the origin only, so you see the assistant’s domain, not the conversation or the question that produced the link.

One wrinkle runs the other way. In session-scoped reports, Google Analytics attributes a session that starts with a direct entrance to the campaign values it already holds for that user (scopes of traffic-source dimensions). So a buyer who first arrived from an assistant and later typed your address can still be counted under the assistant. Check which scope a report uses before you compare one channel’s numbers across reports.

Where do AI Overviews and AI Mode clicks go?

Clicks from AI Overviews and Google AI Mode go to your search numbers. Google Search Central says their traffic is included in Search Console’s overall search traffic, in the Performance report within the “Web” search type (AI features and your website, checked 14 November 2026). In Google Analytics, the AI Assistant channel excludes them, so they arrive with the rest of your Google search clicks.

Neither tool separates them. If Google’s AI features matter in your category, you can’t report their clicks on their own; you can report your search clicks and measure separately how often AI Mode names you.

How do you set up a channel you can trust?

Set up a channel you can trust by writing its definition yourself, so it covers the assistants you care about and stays the same from month to month. Google’s list behind the default channel isn’t published, so you can’t see when it changes. A custom channel group fixes that, and Google says custom channel groups apply to your reports retroactively.

  1. Check the default first. In Reports, open Acquisition, then Traffic acquisition, and find the AI Assistant row. Then set the dimension to Session source and search for assistant domains, to see which ones arrive as Referral instead.
  2. Copy the default group. In Admin, under Data display, open Channel groups and create a new group, which starts as a copy of the default (custom channel groups). A standard property allows two custom groups.
  3. Add your AI channel with a regex. Add a channel whose source “matches regex” a list of the assistant domains you see in your own data. The pattern below is an example to check, not a verified list.
  4. Move it above Referral. Google says traffic joins the first channel it matches, in the group’s order, and its own example tells you to place the AI channel above Referral.
  5. Read it by landing page and key event. Open the Landing page report, add Session source / medium as a secondary dimension, and look at the key events from AI visits on each page.

An example pattern, built from domains to verify against your own Session source values:

^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|www\.perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com)$

Anchor the pattern and list full hostnames, copied exactly as the report shows them. A loose pattern such as .*google.* catches every Google search click, because Gemini’s address contains “google”, and .*gpt.* can catch unrelated sites. When an assistant changes its address, its visits fall outside your channel until you add the new hostname.

What should you report, and next to what?

Report AI referral visits as a floor, with their trend, next to a measure of how often engines name you. On their own, the visits answer “how many people clicked through from an assistant”; they say nothing about the buyers who read an answer and came another way.

Three views make the number useful:

  • Landing pages. Which pages AI visits land on shows which pages the engines link to. Pages that never appear are worth checking against the questions your buyers ask.
  • Key events. Sign-ups and demo requests from the AI channel, and paths with an AI visit on them, which the glossary calls an assisted conversion.
  • Direct, beside it. A rise in Direct sessions to deep pages, which people rarely type, in the same weeks as a rise in AI referrals, is a pattern worth noting. It is a correlation, and it stays one.

Write every change as observed. “AI referrals rose from [a] to [b] sessions since October, in the same weeks we published two comparison pages” holds up; “the pages drove AI traffic” doesn’t, because engines, competitors and seasons changed too. The glossary entry on observed versus caused gives the rule in one line.

The weekly brief has a place for these numbers under the visibility rate. Download the weekly brief template and add one line: AI referral visits this week, with last week’s beside them.

What can’t analytics see?

Analytics can’t see answers that name you without a click, crawler visits it filters out, and the reason any visit happened. Each gap needs a different source:

  • Unclicked answers. The only way to see them is to ask the engines your buyers’ questions and count how often you are named, with a range. That is the visibility rate, and it is the number that leads the brief.
  • Crawlers. Google Analytics says traffic from known bots and spiders is excluded automatically, that you can’t turn this off, and that you can’t see how much was excluded (known bot-traffic exclusion). Your server or CDN logs record those requests. Read them by user agent to tell an engine’s crawler fetching a page from a person visiting it; should you block AI crawlers? lists which crawler does what.
  • Intent. A free-text “how did you hear about us?” field on your demo form catches the buyers who say an AI tool sent them, in their own words.

Where to start

This week, open Traffic acquisition, note last month’s AI Assistant sessions, and list the assistant domains that appear as Referral instead. If any do, build the custom channel group and move it above Referral, then report the count as a floor next to your visibility rate. AmpliRank measures that rate on unbranded questions with its range; the sample report shows one.

Sources

■ Ask · Measure · Compare · Trace · Fix · Share

Measure what AI says about you.

Cookies