Is ChatGPT Sending You Customers? See AI Visitors in GA4 (2026)

Is ChatGPT actually sending anyone to your website? Let me tell you what is almost certainly happening there at this very moment, whether you can see it or not. Someone, somewhere, asked ChatGPT or Perplexity a question, perhaps "who is the best [your trade] in [your town]"; the assistant named you, they clicked through, and they landed on your site. This is the great shift everyone in marketing has been talking about for two years, arriving in your own analytics. And yet, when you open Google Analytics 4 and go looking for it, you find almost nothing. No line called "ChatGPT." No neat little AI category. The traffic seems not to exist.

But it does exist; it is simply hiding, and understanding where it hides is the key to everything that follows. The truth is that GA4, out of the box, was never designed to recognise this new source, and so it quietly files the great majority of your AI visitors under the least helpful label imaginable: "Direct." In this piece we will look at exactly why that happens, how to build a proper AI channel so you can finally see this traffic, and, the honest part that many guides skip, what these numbers can and cannot tell you. Because measuring something badly is often worse than not measuring it at all.

In short: AI referrals hide in GA4 for two reasons. First, much AI traffic comes from apps (people using the ChatGPT app on their phone), and app traffic frequently arrives with no referrer, so GA4 files it as "Direct." Second, even the web referrals that do arrive land in the generic "Referral" channel, un-labelled as AI. The fix is to build a Custom Channel Group in GA4 with a rule that matches AI sources by a regular expression, placed above Referral; from then on, ChatGPT, Perplexity, Gemini, Claude and Copilot appear on their own line. Be honest about the limits: this recovers the referred traffic, not the app traffic that arrived anonymously, so treat the figure as a floor. If the whole idea of AI search is new to you, start with our plain-language guide to what generative engine optimization means for business owners.

Why does AI traffic hide in "Direct"?

To fix a thing, one must first understand it, so let us be precise about the two mechanisms at work here, because they are genuinely different problems.

The first, and larger, is the app problem. A great many people now use AI assistants inside a dedicated mobile app rather than a browser tab. When you tap a link inside a phone app, that click often arrives at the destination website with no "referrer" attached; the little piece of information that tells the website where the visitor came from is simply missing. GA4, faced with a visitor who appears from nowhere, does the only thing it can: it labels them "Direct," the same bucket used for someone who typed your web address from memory. This is not a bug you can fully fix, because the information genuinely never arrived. It is the nature of app traffic.

The second is the labelling problem, and this one you can fix. When AI traffic does arrive from a web browser (someone using ChatGPT or Perplexity in a browser tab) it usually carries a proper referrer, such as chatgpt.com or perplexity.ai. But GA4, by default, does not know these are AI sources. It dutifully files them under the generic "Referral" channel, mixed in with links from every other website, uncounted and uncelebrated as AI. So even the traffic that is trackable is invisible until you teach GA4 to recognise it. That teaching is exactly what we are about to do.

How do you teach GA4 to show AI visitors on their own line?

Here is the good news: solving the labelling problem is genuinely straightforward, and it takes a few minutes. Before I give the steps, one important note on terminology and menus: Google reorganises the GA4 interface with some regularity, so if a label below has moved slightly by the time you read this, look for the nearest equivalent; the concept is stable even when the wording drifts.

The tool you want is called a Custom Channel Group. Follow these steps:

In GA4, click the gear icon to open the Admin area.
Under the Data display section, click Channel groups.
Click Create new channel group and give it a clear name, such as "AI Search".
Add a new channel within it (call it "AI Traffic") and set the condition: Source matches regex, then paste your AI regular expression (below).
Drag this new "AI Traffic" channel above the existing "Referral" channel, so AI visitors are caught by your rule before they fall into the generic Referral bucket. Order matters here.
Save the channel group.

For the regular expression, the pattern that tells GA4 which sources count as AI, a reliable, current version covering the major engines is:

You may extend it as new engines become meaningful; the landscape shifts, and it is wise to revisit the pattern every few months to add newcomers that have grown from negligible to measurable. From the moment you save, your reports will begin showing AI as its own channel, separate from Referral, so you can finally watch it grow.

Does this fix the past, or only the future?

This is where I must be careful and honest, because it is a point that trips people up. A Custom Channel Group is a reporting layer; it reorganises how GA4 displays the data it already holds; it does not alter or rewrite the underlying data. This has two consequences worth understanding.

On the encouraging side: because it is reorganising existing data, when you view reports through this channel group, sessions that already occurred and carried an AI referrer can be shown under your new AI channel. In that sense the grouping does apply to your historical sessions as you look back through it.

On the sobering side: it can only surface what was genuinely recorded. The app traffic that arrived with no referrer was filed as "Direct" at the time, with no AI fingerprint attached; no reporting layer can retroactively invent information that was never captured. So your AI channel will show you the referred AI traffic, but the app-driven portion remains lost in Direct. This is why I keep urging you to treat the number as a floor, a minimum. Your true AI traffic is somewhat higher than the figure your beautiful new channel reports. Better an honest floor than a false certainty.

What should you actually measure once you can see it?

Seeing the traffic is only the beginning; the point is to learn from it. Once your AI channel is live, here is what I would watch:

Watch the trend far more than the absolute number. Because of the Direct-bucket problem, the absolute figure understates reality, but the direction of travel, month over month, is honest and revealing. Is AI traffic growing? Accelerating? That trend is your real signal.

Watch which pages AI visitors land on. This tells you which of your pages the engines are actually citing and sending people to, invaluable intelligence for deciding where to invest. If Perplexity keeps sending people to one particular guide, that guide is doing something right, and you should learn from it.

Watch what they do next: do AI visitors book, call, submit a form? Set up conversions in GA4 so you can see whether this traffic merely arrives or actually turns into customers. Traffic that does not convert is a vanity metric; traffic that converts is a business.

And, alongside GA4, keep doing the manual check that no analytics tool can replace: periodically ask the AI engines your customers' real questions and see whether you appear. GA4 tells you about the visitors who came; the manual check tells you whether you are even in the running.

A story about the traffic that was there all along

I spoke with a business owner who was, frankly, dismissive about the whole AI-search phenomenon. "It's hype," he told me. "I've checked my analytics; I get nothing from ChatGPT." And in one narrow sense he was right: his GA4 showed a flat, empty nothing where AI should have been. He had looked, seen zero, and concluded the world was exaggerating.

So we built him a Custom Channel Group, exactly as described above, and asked him to leave it running for a few weeks. When we looked again, there it was: a modest but unmistakable and growing line of AI traffic that had, all this time, been sitting inside his "Direct" and "Referral" buckets, uncounted. And here is the part that changed his mind entirely: several of those AI visitors had gone on to fill out his contact form. Real enquiries, from real people, sent by a machine he had insisted was hype. He had not been wrong that his reports showed nothing; he had been wrong to trust a report that was never set up to show him the truth. The traffic had been there all along. He simply had never taught his tools to see it. It is remarkable how much of the AI shift is exactly this: real, arriving, and invisible only because no one looked properly.

If reading this has made you suspect you are flying blind, that you genuinely do not know how much of your business now originates in an AI conversation, our SEO Standings check shows, in plain language, where you currently stand across both Google and the AI engines, and for owners who would rather have the whole measurement-and-visibility effort handled properly, our done-for-you SEO service exists for exactly that.

Frequently Asked Questions

Why does GA4 show ChatGPT traffic as "Direct" instead of a referral?

Chiefly because a large share of AI traffic comes from mobile apps, and links tapped inside an app often arrive at your site with no referrer information attached. When GA4 receives a visitor with no source information, it defaults to labelling them "Direct," the same category as someone typing your address from memory. Web-browser AI traffic usually does carry a referrer, but GA4 files it under the generic "Referral" channel unless you build a custom rule to identify it as AI.

How do I create an AI traffic channel in GA4?

In GA4, open Admin (the gear icon), go to the Data display section, and select Channel groups, then Create new channel group. Add a channel called something like "AI Traffic" with the condition "Source matches regex" and paste a regular expression listing the AI domains (chatgpt.com, perplexity.ai, gemini.google.com, and so on). Crucially, drag this channel above the existing Referral channel so AI visitors are caught by your rule first, then save. GA4 does rename its menus periodically, so if a label has moved, look for the nearest equivalent.

Will my new AI channel show me past traffic too?

Partly. A custom channel group is a reporting layer that reorganises data GA4 already holds, so when you look back through it, past sessions that carried an AI referrer can appear under your AI channel. However, it cannot recover information that was never recorded; the app traffic that arrived with no referrer was filed as "Direct" at the time and stays there. So the channel gives you an honest view of referred AI traffic, but the true total is somewhat higher; treat the figure as a floor.

What is the most useful thing to measure about AI traffic?

The trend over time, and what those visitors do after they arrive. Because the absolute number understates reality (thanks to the Direct-bucket problem), the month-over-month direction is your most honest signal: is AI traffic growing and accelerating? Beyond that, watch which pages AI visitors land on (this reveals what the engines are citing) and set up conversions so you can see whether this traffic actually turns into calls, bookings, or enquiries rather than merely arriving.

Sources: Orbit Media: how to track AI referral traffic and leads in GA4; why app-based AI traffic is categorised as Direct in GA4 (Orbit Media); building a GA4 custom channel group with an AI regex (Discovered Labs).