In GA4, visits from AI assistants show up as referral traffic with sources such as chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com or claude.ai. To see them together, create a custom channel group with an "AI assistants" channel defined by a regular expression on the source, and place it above Referral. Keep in mind that some of this traffic arrives without a referrer and gets counted as direct, and that clicks from AI Overviews are recorded as Google organic search.
- ChatGPT usually adds utm_source=chatgpt.com to the links it cites, which makes it easier to measure.
- A custom channel group rolls every AI source into a single row in your reports.
- Channel order matters: the AI channel must sit above Referral.
- Clicks from AI Overviews and AI Mode count as google / organic, not as AI.
- Track key events (forms, WhatsApp, purchases), not just sessions.
Why measure traffic from AI assistants
More and more people ask ChatGPT, Perplexity or Gemini directly, "Which auto shop in Torreón can I trust?" or "Who builds websites in La Laguna?" When the assistant cites your site with a link and someone clicks it, that visit lands on your page. If you don't separate it out, it gets lost among "referrals" and you'll never know whether your GEO work is paying off.
Measuring it lets you answer concrete questions: Which pages get visits from AI? Do those visits convert better or worse than visits from Google? Which assistant sends you the most people? With that data, you decide where to invest instead of chasing trends.
It's also useful for having honest conversations with your agency or team. If someone tells you "you're already showing up in ChatGPT," the right question is: how many visits, to which pages, and how many leads did they bring? With the channel set up properly, that answer takes two clicks instead of an argument.
Before you start, confirm the basics: your GA4 property is installed on every page (including guides and blog posts), your cookie banner isn't blocking measurement by mistake, and you have Editor access. Without that, any channel you set up will show incomplete numbers.
What this traffic looks like in GA4
Visits from assistants usually arrive as referral traffic, with the assistant's domain as the source. These are the most common sources as of late 2026:
| Assistant | Source you'll see in GA4 | Notes |
|---|---|---|
| ChatGPT | chatgpt.com (formerly chat.openai.com) | Usually adds utm_source=chatgpt.com to cited links |
| Perplexity | perplexity.ai or www.perplexity.ai | Cites linked sources in almost every answer |
| Gemini | gemini.google.com | Not every visit carries a referrer |
| Microsoft Copilot | copilot.microsoft.com | Some visits may come through bing.com |
| Claude | claude.ai | Smaller volume, but growing in professional use |
| Google AI Overviews and AI Mode | google / organic | Not broken out in GA4; counted as organic |
To see them today without configuring anything, go to Reports > Acquisition > Traffic acquisition, switch the dimension to Session source / medium and look for those domains. If your cited links carry only utm_source, the medium may show as "(not set)"; that's normal.
Step by step: create an "AI assistants" channel
Checking domain by domain gets old fast. The better option is to create a custom channel group that rolls everything up:
- Go to Admin > Data display > Channel groups.
- Choose Create new channel group. GA4 copies the default channels so you don't start from scratch.
- Give it a clear name, such as "Channels with AI."
- Add a new channel called AI assistants with the condition: Source → matches regex → the expression below.
- Reorder the channels and move "AI assistants" above "Referral." GA4 assigns each session to the first channel whose condition it meets; if Referral comes first, it swallows everything.
- Save, and in your acquisition reports, switch the channel group to the one you just created.
A regular expression that works well as a starting point:
.*(chatgpt|openai|perplexity|gemini\.google|copilot|claude\.ai|deepseek|meta\.ai).*
Review it every few months: new assistants appear and some change domains. If you run your own campaigns with a utm_source containing any of those words, adjust it so the data doesn't get mixed.
Common setup mistakes
When we review GA4 accounts for businesses in La Laguna, these are the slip-ups we see most often:
- Channel placed below Referral. This is mistake number one: the channel exists, but it always shows zero because Referral already claimed those sessions.
- An overly broad expression. If you just enter "ai," you'll catch any source that contains those two letters, like an email domain or an ad network. Use full domain names.
- Filtering by "First user source" instead of "Session source." The first tells you how someone arrived the very first time; to measure repeat visits from assistants, use the session dimension.
- Misconfigured referral exclusions. If you added chatgpt.com to your list of unwanted referrals by mistake, GA4 reassigns those sessions and you'll never see them.
- Looking at a single week. AI volume is usually small for local businesses; compare periods of at least a month so you don't draw conclusions from three visits.
Measure what matters: key events
Seeing sessions is nice, but leads and sales are what pay the rent. In GA4, mark the actions that really count as key events (formerly "conversions"):
- Contact or quote form submissions.
- Clicks on the WhatsApp button. If you use a wa.me link, you can create it with our WhatsApp link generator and track the click as an event.
- Clicks on your phone number or on "Get directions."
- Completed purchases, if you have an online store.
Then, in the acquisition report, filter by the "AI assistants" channel and compare its key event rate with organic search. On several sites we measure, visitors who come from an assistant already have their question half-answered and tend to arrive with clear intent; but that varies by industry, so trust your own data, not someone else's averages.
For your own campaigns, tag links with the UTM builder so you don't confuse what you share with what AI recommends on its own.
The limits: what GA4 won't tell you
Being honest about the data prevents bad decisions. Here's what GA4 can't see:
- Traffic without a referrer. Many desktop and mobile assistant apps open links without passing along the source domain. Those visits land in "Direct." If you see direct traffic to an in-depth guide climbing for no apparent reason, some of it is probably coming from AI.
- Mentions without a click. If ChatGPT recommends your business by name and the person then looks you up on Google Maps or messages you on WhatsApp, GA4 doesn't attribute it to AI. Ask "How did you hear about us?" on your form or when you close the sale, and include the option "ChatGPT or another AI recommended you."
- AI Overviews. Clicks from Google's AI answers are recorded as google / organic. In 2026, Google added impression reporting for its generative AI features to Search Console; we explain it in our AI Overviews guide.
- Bots. Crawler visits from bots like GPTBot or OAI-SearchBot don't execute JavaScript, so GA4 doesn't count them. To see them you need your server logs or your CDN dashboard. More about those bots in our robots.txt for AI guide.
A simple dashboard to review every month
You don't need a dashboard with 30 charts. A GA4 exploration or a Looker Studio report with these four views is plenty:
- Sessions by channel using the custom group, month over month.
- Landing pages filtered by "AI assistants": tells you which content gets cited.
- Key events by channel: how many leads or sales AI contributes.
- Source within the channel: which assistant sends you the most visits.
An example from La Laguna: a real estate agency in Torreón (name withheld) discovered that nearly all its visits from assistants landed on a single guide about the home-buying paperwork process, not on its property listings. With that insight, they added a consultation form inside the guide and a WhatsApp call to action. AI visits stayed small compared with Google, but they stopped leaving without a trace.
What to do with what you learn
Data is only useful if it changes decisions. Once you have two or three months of history:
- If a page gets AI visits, strengthen it: update the facts, add a table, an FAQ section and a visible call to action. The recipe is in AI-citable content.
- If nothing comes in, check the basics: make sure your robots.txt doesn't block OAI-SearchBot or PerplexityBot and that your pages answer real questions. The SEO + GEO audit gives you a diagnosis in seconds.
- If the AI channel converts well, consider producing more guides like the one that works, focused on your customers' questions.
If you'd like us to set up GA4, key events and the dashboard for you, we do it as part of our SEO for AI or digital marketing services.
