See how much traffic ChatGPT and Perplexity send you in GA4: the AI Assistant channel, what it misses, and a custom channel group

GA4's default channel group now has an AI Assistant channel. Google's help page says it's for visits from sources like ChatGPT, Gemini, DeepSeek, Copilot and Grok, and that it excludes Google's AI Overviews and AI Mode. What the page doesn't do is print the full list of sources.

That matters the moment someone asks you "how much traffic do we get from AI?" If Perplexity visits are still sitting in Referral, the built-in number is low, and you won't know unless you check.

This guide is the 30-minute check I'd run on any site before I report AI traffic to a team. You check what the built-in channel catches, build a custom channel group with one regex, read the traffic by landing page and write down what GA4 can't see. It pairs with the tracking spreadsheet, which measures whether AI answers name you. This one measures who clicks through.

What you'll have when you're done

• A list of AI sources in your property and which channel each one lands in today
• A custom channel group with an AI channel that sits above Referral
• A landing-page view of AI traffic, so you know which pages the assistants send people to
• A short monthly note for your team with the number and its known gaps
• [SCREENSHOT: GA4 Traffic acquisition with the custom AI channel row and its sessions]

Before you start

• A GA4 property with at least a few weeks of data, and Editor access (custom channel groups live in Admin).
• Free. Nothing here needs a paid tool. Looker Studio is optional for the monthly view.
• About 30 minutes.

Google's pages for this: the default channel group definitions and custom channel groups.

Step 1: Find every AI source already in your data

Open Reports > Acquisition > Traffic acquisition. Add Session source as a secondary dimension, set the date range to the last 90 days and search the table for each of these, one at a time.

Search terms for the source column
chatgpt   openai   perplexity   gemini   copilot
claude    deepseek   grok   you.com   meta.ai

For each hit, write down the source and the channel it's in. Some of these sources may not exist in your data yet, and that's fine.

Step 2: Copy the default group into a custom one

Go to Admin > Data display > Channel groups and create a new channel group by copying the default. Name it something your team will understand, like "Channels + AI".

A standard GA4 property gets two custom channel groups on top of the default, and each one can hold up to 50 channels, so don't burn both on experiments.

Step 3: Add one AI channel, above Referral

Add a new channel called AI tools. Set its condition to Session source matches regex and paste this:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|chat\.deepseek\.com|grok\.com|you\.com|meta\.ai

Add any other source you found in Step 1. Escape the dots, and don't shorten it to something like .*ai.*, which would also catch email and half the web.

Then move AI tools above Referral. Google's help page says traffic is assigned to the first channel whose definition it matches, so a channel sitting below Referral never gets anything. Save.

Step 4: Read AI traffic by landing page

I think the total is the least useful number here. What helps is which pages the assistants send people to. Open Reports > Engagement > Landing page, add your new channel group's channel as a secondary dimension and filter to AI tools (or AI Assistant if you skipped the custom group).

Look at three things for each page: sessions, engaged sessions and key events.

AI landing pages (one row per page per month)
page | ai sessions | engaged sessions | key events | cited in spreadsheet? (y/n)

The last column links this to the tracking spreadsheet. A page that answers get cited from and that also gets clicks is the page to protect and update first. The pages AI answers cite guide covers what those pages tend to look like.

Step 5: Write down what GA4 can't see

AI traffic in GA4 is a floor, not a total, and I'd say so in the report every time. Three gaps come up on almost every site.

App visits with no referrer. Clicks from mobile and desktop apps often arrive without a referrer and land in Direct. ChatGPT often adds utm_source=chatgpt.com to the links it shows, which helps, but don't count on every assistant doing that.

AI Overviews and AI Mode. Google's help page says these are excluded from the AI Assistant channel. The clicks come from google.com, so they count as Organic Search and you can't split them out in GA4.

Visibility without clicks. An answer can name you and send nobody. That's what the spreadsheet and the crawler log check are for.

Monthly AI traffic note
AI sessions this month: [N] ([+/- %] vs last month)
Top 3 AI landing pages: [page], [page], [page]
Sources: [chatgpt.com N], [perplexity.ai N], [other N]
Not counted: app visits with no referrer (in Direct),
AI Overview / AI Mode clicks (in Organic Search)

Step 6: Check it again in a month

Re-run Step 1 once a month. New assistants show up as new sources, and Google can change what the built-in channel covers. When a new one appears in Referral, add it to the regex and save. The custom group re-applies to past data, so last month's report fixes itself.

What I'd do differently

Check before you build. I expected to need the custom group everywhere. If the built-in channel already catches every AI source in your data, you don't.

Report pages, not just a total. The landing-page view is what changes what the team writes next.

Put the gaps in the report. A small AI number with its blind spots written next to it is more useful than a bigger number nobody can defend.