Inside Google's search results, "AI Mode" — where an AI answers your question directly — has been spreading. The natural question is: when someone arrives from AI Mode, which channel do they show up under in GA4 (Google Analytics 4)? It's tempting to assume "they should all group under some AI-related channel," but it isn't that simple. AI Mode is a feature inside Google Search, so a visit that comes from it is treated as having Google as its referrer, and it's likely to land in Organic Search — or in Direct — rather than in an AI-specific channel. This article lays out where AI Mode traffic falls in GA4, and why reading "the AI-channel number equals all of AI" makes you undercount revenue, in plain terms as of July 2026. The specifics can still change, so check the latest official information before you make a call.
Table of contents
This article in brief#
-
AI Mode is an AI feature inside Google Search, so its referrer is Google
A visit that arrives from an AI Mode answer is treated as having Google as its referrer, so in GA4 it's likely to land in Organic Search
-
That's why it rarely surfaces in an AI-specific channel, and scatters across several buckets
When the referrer isn't carried over, it falls into Direct too. It doesn't necessarily gather into a channel you can identify as AI
-
Reading "the AI channel is small, so AI isn't working" makes you lose revenue
Judge only by the traffic you can identify as AI by referrer and the AI-Mode revenue hidden in Organic Search drops out entirely
-
Don't watch the size of a single label — tie AI traffic to revenue
Connect the AI traffic you can see through to revenue (RPS), and honestly treat the part you can't isolate as "not yet visible"
What AI Mode is#
The bottom line: AI Mode is a feature where, inside Google Search, an AI assembles and returns an answer directly, and a visit coming from it has Google as its referrer.
Search used to work like this: you type a keyword, a list of pages lines up, and you pick a link yourself. In AI Mode, the AI answers your question on Google's search surface, with reference links tucked into that answer. The visitor reads the AI's answer first and, if needed, comes to a store's site through a link inside it. The entry point is still Google Search, but now there's a layer of AI answer in between.
This shift in how things look is tied to measurement too. On June 26, 2026, a report covering how generative AI surfaces your site became available in Google Search Console in Japan as well. But what you mainly learn there is impressions on the search surface — how many times the link was actually clicked, and how much revenue it produced, is not something that report alone can trace. The blind spot between impressions and revenue is covered in detail in Why the GSC generative-AI report doesn't show revenue.
"Fine, we know the impressions. So where do the people who clicked through show up in GA4?" That's the heart of this article. To say the conclusion first: AI Mode visits don't come out in an AI-specific channel — they surface mixed into Organic Search and Direct.
Where AI Mode traffic shows up in GA4#
The bottom line: because the referrer is Google, AI Mode visits are likely to land in Organic Search, and when the referrer isn't carried over, they fall into Direct too. They rarely surface in a dedicated channel you can tell apart as AI.
First, the premise. Traffic like ChatGPT or Perplexity — where the referrer tells you it's an AI tool — can be grouped into an AI-related channel on the GA4 side. The caveats to that view are covered in Blind spots when reading GA4's AI Assistant traffic. What this article calls the "AI channel" for convenience is that bucket of traffic identifiable as AI by referrer.
Google's AI Mode is a different story. Because AI Mode is a feature inside the single service that is Google Search, a visit that comes through it arrives with google as the referrer. GA4 sorts traffic with a Google referrer into Organic Search, so AI-Mode visits are also likely to land in Organic Search, indistinguishable at a glance from ordinary Google search. The official rules for how channels get sorted are laid out in Decoding GA4's channel classification from the official definitions.
On top of that, when the referrer information isn't carried over cleanly, the destination shifts to Direct. The mechanism by which AI-related traffic slips into Direct is covered in detail in Why AI traffic hides in Direct. For AI Mode too, visits with a missing referrer fall into Direct, so the buckets end up split across two — Organic Search and Direct.

What matters here is not to declare "AI Mode is always classified this way." How the referrer is passed varies with your site's settings and the browser's behavior, and Google's and GA4's specs move too. What is dependable is only the direction: as long as AI Mode is a feature inside Google Search, it tends to scatter across several buckets like Organic Search and Direct rather than gather neatly into a dedicated AI channel. You can't state it as a certainty, but keeping this tendency in mind is enough to avoid the next misread.
The misread that the AI channel is small#
The bottom line: if you look only at the "AI channel" you can identify by referrer and conclude "AI is still small, so it isn't working," the AI-Mode revenue mixed into Organic Search drops out entirely.
Say GA4 shows "120 visits last month" in the AI channel. Looking at that number alone, you're tempted to think "AI is still just a few percent of the total, so it can wait." But behind it, the people who came from Google's AI Mode are mixed into Organic Search, not the AI channel. In other words, outside those visible 120 visits, there may be a separate set of AI-Mode visits that aren't being counted.
What makes this leak nasty is that the revenue sold through AI Mode gets added to Organic Search revenue too, and never stands up as "AI-derived revenue." Mistake the AI-channel number for the whole picture, and you'll estimate AI's contribution as smaller than it really is and misjudge your investment. For an e-commerce operator, this is exactly where the blind spot sits.
In fact, the spot where many operators trip up around AI traffic is that "measurement is the hardest part." Even if traffic from ChatGPT or Perplexity is somewhat visible, why that page was chosen — and how much the whole picture, including Google's in-search AI, generated in revenue — is hard to read off the standard reports.

Put differently, judging AI's effect by "the size of a label called the AI channel" is shaky to begin with. The label grows or shrinks depending on how the buckets happen to fall. What you should watch are two things: how much the AI traffic you can see is actually selling, and how honestly you treat the part you can't see. That thinking leads into the next solution.
RevenueScope — how it helps
The bottom line: showing traffic per AI engine that's visible by referrer as its own row, with revenue and RPS, while honestly disclosing the part you can't isolate as "undercounted," is the starting point for tracking AI traffic by revenue.
RevenueScope doesn't bury AI traffic into one lump. It splits it by engine — ChatGPT, Claude, Perplexity, Gemini, Copilot — and shows each engine's visits, revenue, and RPS (revenue per visit), already computed. From the referrer and access characteristics, it tells human visits apart from bots (automated access) and displays clean numbers with bots removed. Why it's worth seeing revenue per AI engine is explored further in Seeing AI-engine traffic and revenue separately.
For example, traffic and revenue by AI engine come back like this (the display uses sample data from a fictional site).
| AI engine | Visits | Revenue | RPS |
|---|---|---|---|
| ChatGPT | 95 | ¥15,945 | ¥167 |
| Perplexity | 64 | ¥11,510 | ¥179 |
| Gemini | 33 | ¥33,188 | ¥1,005 |
| Claude | 10 | ¥12,088 | ¥1,208 |
| Copilot | 12 | ¥9,424 | ¥785 |
Even when the traffic looks small by visit count, RPS can reveal that engines like Gemini or Claude earn strikingly more per visit. Precisely because you're viewing by revenue efficiency rather than visit volume, you can notice differences like these.

But here is the crucial point we have to be honest about. What appears in this table is only the traffic you can identify as an AI tool by referrer. Because Google's AI Mode arrives with google as the referrer, it does not show up in these per-AI-engine rows. On the same sample site, Google Search has 378 visits (¥128,459) and Direct has 226 visits (¥137,253), and AI-Mode visits are mixed into these. Even RevenueScope cannot fully and cleanly isolate Google's in-search AI. Because the AI classification is based on the referrer, an AI that doesn't pass a referrer will be undercounted, and the correct reading is that the actual AI traffic is larger than the numbers shown here (it comes out undercounted).
The value of RevenueScope is not to claim it can capture AI Mode perfectly. It's that it brings "AI-engine traffic × revenue" — which is structurally hard to surface in GA4's standard reports — onto one screen with clean, bot-excluded numbers, and, on top of that, doesn't hide the part it can't isolate. Rather than being tossed around by the size of a single label, connect what you can see through to revenue, and honestly treat what you can't see as "not yet visible." Only when those two come together can you use AI traffic as decision material for revenue.
FAQ#
Q1. Which GA4 channel do people from AI Mode show up in?#
Because the referrer arrives as Google, they're likely to land in Organic Search. When the referrer information isn't carried over, they fall into Direct too. They rarely surface in a dedicated channel you can tell apart as AI, so assume the buckets scatter across several. That said, it can change with your site's settings and browser behavior, so it isn't something you can say will always be classified this way.
Q2. Can RevenueScope fully isolate AI Mode traffic?#
No, not completely. Traffic identifiable as AI by referrer, like ChatGPT or Perplexity, can be shown as its own row per engine with revenue and RPS. But Google's AI Mode arrives with google as the referrer, so it's mixed into Google Search (Organic Search) rather than the per-AI-engine rows. That's exactly why we connect what's visible through to revenue and honestly disclose the part we can't isolate as undercounted.
Q3. While the AI channel is still small, can I ignore it?#
Not recommended. Even if the AI channel visible by referrer is small, AI-Mode visits and revenue may be mixed into the Organic Search outside it. Rather than deciding "it isn't working" from the size of a single label, build the habit of connecting the AI traffic you can see through to revenue (RPS), and you won't misjudge when AI traffic grows.
Wrap-up#
- AI Mode is an AI feature inside Google Search, so the referrer of a visit coming from it is Google
- Because of that, in GA4 it's likely to land in Organic Search, and when the referrer is missing, in Direct too. It rarely surfaces in an AI-specific channel
- Judge only by the size of the "AI channel" you can identify by referrer, reading "AI is small, so it isn't working," and you lose the AI-Mode revenue mixed into Organic Search
- What to watch is not the size of a single label, but the revenue of the AI traffic you can see and an honest stance toward the part you can't isolate
See which ads actually drive revenue, at a glance
Free up to 5,000 sessions/month, AI analyst included. No credit card required. Up and running in 5 minutes.





