How to track AI traffic in GA4 from ChatGPT, Perplexity and more

To track AI traffic in GA4, create a custom channel group with an "AI assistants" channel that matches sources such as chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com, and place it above Referral. That one change turns a scattered list of referrers into a channel you can trend, segment and tie to conversions.

It matters because leadership teams are asking a fair question: is AI search sending us customers, or only taking clicks away? Without a clean view, the honest answer is "we don't know". Visits from ChatGPT sit in the Referral report beside directories and partner sites. Visits from apps that strip the referrer fall into Direct. Clicks from Google AI Overviews look identical to any other Google organic visit.

This guide walks through the GA4 setup, the sources to include, what GA4 cannot see and how to fill that gap, and a simple monthly report that connects AI visibility to pipeline. It assumes a standard GA4 property and editor access, and nothing more.

Why GA4 hides AI traffic by default

GA4's default channel group has no rule for AI assistants, so their visits are filed under Referral, Direct or Organic Search depending on how the click arrived. Each of those paths needs different handling.

• Referral. When someone clicks a cited link in ChatGPT, Perplexity, Claude, Gemini or Copilot on the web, the browser usually passes the assistant's domain as the referrer. GA4 records it as a referral source, mixed in with every other site that links to you.
• Direct. Desktop and mobile apps, copied links and some privacy settings remove the referrer. Those visits show up as direct traffic with no clue where they came from.
• Organic Search. Clicks from Google AI Overviews and AI Mode come from Google itself, so GA4 classifies them as google / organic. There is no parameter that separates them from a click on a classic blue link.

The result is that most companies undercount AI traffic, and the part they can see is buried. Fixing the referral piece is quick. The other two need a different approach, covered further down.

Step 1: Find the AI sources already in your data

Before building anything, confirm which assistants are already sending visits. Open Reports, then Acquisition, then Traffic acquisition. Change the primary dimension to Session source and search the table for "chatgpt", "perplexity", "claude", "gemini" and "copilot".

You will often see more than one variant per product. ChatGPT, for example, can show up as chatgpt.com, as the older chat.openai.com, or as a source taken from a utm_source=chatgpt.com parameter that ChatGPT adds to many of the links it cites. Write down every variant you find, along with any smaller assistants your audience uses, such as DeepSeek, Mistral's Le Chat, You.com or Meta AI. This list becomes your regex.

Step 2: Track AI traffic in GA4 with a custom channel group

A custom channel group lets you add an "AI assistants" channel without touching the default channel group that your existing reports rely on. Google documents the feature in its custom channel groups help article. The setup takes about fifteen minutes.

1. Go to Admin, then Data display, then Channel groups, and choose Create new channel group. If you prefer, copy the default group first so every standard channel is already in place.
2. Name it something clear, such as "Channels with AI".
3. Add a new channel called "AI assistants".
4. Set the condition to Source, matches regex, and paste a pattern like the one below.
5. Reorder the channels so "AI assistants" sits above Referral. GA4 assigns each session to the first channel whose rule it matches, so if Referral comes first it will claim these visits.
6. Save. The group applies to past data as well as new sessions, so you can compare previous months straight away.

A starting regex that covers the main assistants:

.*(chatgpt|openai|perplexity|claude\.ai|gemini\.google|copilot\.microsoft|deepseek|chat\.mistral|you\.com|meta\.ai).*

The leading and trailing wildcards make it tolerant of subdomains and app variants. Keep "gemini.google" and "copilot.microsoft" specific so you do not catch unrelated sites that happen to contain those words. Then test it: open Traffic acquisition, switch the dimension to your new channel group and confirm the AI assistants row matches the sources you listed in step 1.

Optional: a saved exploration for deeper analysis

For landing-page and conversion analysis, build a free-form exploration with Session source as a row, Landing page as a second row, and Sessions, Engaged sessions, Key events and Session key event rate as values. Filter it to your AI channel. This shows which pages assistants send people to, which is usually a short and revealing list.

Step 3: Measure what GA4 cannot see

Referral tracking captures only part of AI influence, so add two more signals: Google's own data for AI Overviews, and self-reported attribution for dark AI traffic.

AI Overviews and AI Mode

Google states in its documentation on AI features that clicks and impressions from AI Overviews and AI Mode are included in the Search Console performance report. They are not split out, so you track them indirectly:

• Keep a list of your priority queries that trigger an AI Overview and check monthly whether you are cited.
• Watch click-through rate on those queries in Search Console. A stable position with a falling CTR usually means an AI answer is absorbing the click.
• Pair this with citation monitoring. Our guide on how Google AI Overviews choose sources covers what earns a citation in the first place.

Dark AI traffic

A buyer who reads a recommendation in the ChatGPT app and then types your brand name into a browser will show up as direct or branded organic traffic. Analytics cannot attribute that visit. Two cheap fixes help:

• Add a "How did you hear about us?" field to demo, contact and quote forms, with "ChatGPT or another AI assistant" as an option. Sales teams are often surprised by how often it gets picked.
• Watch branded search and direct traffic alongside your AI channel. If AI mentions rise and branded demand rises with them, that is a reasonable signal even without click-level proof.

Server logs

If you have access to raw logs, look for user agents such as ChatGPT-User and Perplexity-User. These are fetches triggered when someone asks an assistant a question that needs your page, which tells you which URLs assistants are reading on behalf of real users, even when nobody clicks through.

Step 4: Connect AI traffic to conversions and pipeline

Session counts alone undersell AI traffic; the useful number is how often those sessions turn into leads or revenue. Make sure your key events (form submissions, demo requests, purchases) are configured in GA4, then compare the AI assistants channel against Organic Search and Referral on session key event rate.

If you run a CRM, pass the session source into a hidden form field so the original channel travels with each lead. That lets you report on AI-sourced pipeline and closed revenue, not only on sessions. It is the same plumbing you would use for paid campaigns, pointed at a new channel, and the same revenue-first reporting we set up in our AI SEO services.

What to put in a monthly AI traffic report

Keep the report short enough that leadership reads it: one table, three trends and a note on what you will do next.

Metric	Where it comes from	What it tells you
AI assistant sessions by source	GA4 custom channel group	Which assistants send visits, and the trend
Key event rate for AI sessions	GA4	Whether AI visitors convert better or worse than other channels
Top landing pages from AI	GA4 exploration	Which pages assistants trust enough to cite
AI-sourced leads and pipeline	CRM with source field	Commercial impact
Self-reported AI attribution	Form field	Influence that analytics cannot see
Share of model	Prompt tracking	How often you are mentioned, clicked or not

The last row matters. Traffic tells you what happened after a click; share of model tells you how often you were in the answer at all. A brand can be recommended constantly and still see modest clicks, because many users act on the answer later. Reporting both prevents the wrong conclusion.

Common tracking mistakes

1. Placing the AI channel below Referral. The channel exists but stays empty because Referral claims every session first.
2. Editing the default channel group. You cannot change it, and copying it carelessly can break comparisons. Keep a separate custom group.
3. Forgetting UTM-tagged visits. Matching only the referrer misses sessions where ChatGPT's own utm_source parameter set the source.
4. Judging AI on volume alone. A small channel with a high conversion rate can be worth more than a large one with a low rate.
5. Never updating the regex. New assistants and new domains appear regularly. Review the Referral report every quarter.

Turning measurement into growth

Tracking is the starting point, not the result. Once you know how to track AI traffic, the next question is how to get more of it: which prompts mention competitors but not you, which pages assistants already cite, and which gaps are worth closing first.

• Build a custom channel group with an AI assistants channel above Referral.
• Use Search Console and citation checks for AI Overviews, since GA4 cannot separate them.
• Add a self-reported attribution option to capture dark AI traffic.
• Report AI sessions, conversions and share of model together.

To see where you stand today, run a few buyer prompts through our free AI Visibility Checker, or read how to get your brand recommended by ChatGPT. If you would like a second pair of eyes on your setup, our free AI SEO audit includes a review of how AI traffic is being measured, and our answer engine optimization service covers the work of growing it.

Frequently asked questions

Does GA4 track ChatGPT traffic automatically?

GA4 records visits from ChatGPT, but by default they land in the Referral channel next to every other website, or in Direct when no referrer is passed. You need a custom channel group or a saved comparison to see AI assistants as their own line.

Can I see Google AI Overview clicks in GA4?

No. Clicks from AI Overviews and AI Mode arrive as ordinary Google organic visits, and Google includes them in the Search Console performance report without a separate filter. Track whether you are cited with prompt and query monitoring, and watch click-through rate on queries that show an AI Overview.

Why is my AI traffic so small compared with organic?

For most sites AI assistants send far fewer visits than Google, and part of what they do send is recorded as Direct. The visits that do arrive often convert well because the visitor has already been given a shortlist. Judge the channel on pipeline per session as well as volume.

Will a custom channel group change my historical data?

Custom channel groups in GA4 apply retroactively to reports, so you can compare past months once the group is saved. Your default channel group stays untouched, which keeps existing dashboards stable.

Which AI sources should the regex include?

At minimum ChatGPT, Perplexity, Claude, Gemini and Microsoft Copilot. Review your Referral report every quarter and add any new assistant domains that appear, since the list of AI products changes quickly.