You have already worked for to appear in ChatGPT and in other AIs. The next question is obvious: how many visits are they sending you?
You open Google Analytics and find a clue: the new channel «AI Assistant», with a handful of sessions. Is that all?
No. That number is just the tip of the iceberg: most of the AI-influenced visits are hidden in other channels.
In this guide, we show you what the AI Assistant channel measures (and what it doesn’t), how to see the rest in 5 steps, and what part of the traffic you will never be able to see (and how to estimate it).
The «AI Assistant» channel: what it is and why it falls short
Since May 13, 2026, Google Analytics classifies the traffic from the AI assistants it recognizes into its own channel: AI Assistant.
You can see it in Reports → Acquisition → Traffic Acquisition, alongside Paid Social, Organic Search, and the rest:
Technically, when Google recognizes the referrer of an AI, it assigns the medium ai-assistant and groups those sessions into the channel.
It’s a huge step: AI traffic finally officially exists for Google. But it has three major limitations:
- It mixes all AIs into a single number. You don’t know how much came from ChatGPT and how much from Gemini or Perplexity without looking at the source of each session.
- The list of sources is from Google and they don’t publish it completely. When it was launched, it named ChatGPT, Gemini, and Claude; weeks later the documentation listed Deepseek, Copilot, and Grok, and Claude had disappeared. You don’t control what goes in and what doesn’t.
- It only counts forward. Your AI traffic prior to May 2026 remains buried in Referral, and visits without referrer (apps, copied links) continue to fall into Direct.
That’s why the dedicated channel of «AI Traffic» that we set up in this guide is still worth it: you control the list, see each AI separately, and can look at the history.
Why is most AI traffic still hidden?
Because the default channels of Google Analytics (the current version, GA4) were defined before AIs sent traffic, and the AI Assistant channel patch doesn’t cover everything.
When someone clicks on a link within an AI, the visit comes with the source of that site: chatgpt.com, gemini.google.com, perplexity.ai.
Google Analytics doesn’t recognize it as anything special and throws it into the «Referral» channel, mixed with any other site that links to you.
Since mid-2025, ChatGPT also adds its own tag to the links: utm_source=chatgpt.com. That helps, but it’s not enough: without your own configuration, you still lack a channel that groups all AI traffic.
And be careful with this: mobile apps of AIs and users of free versions often do not pass the source data. Those visits fall into «Direct», completely invisible.
The data that justifies it: this traffic converts 4.4 times more
Before touching configuration, the why.
Traffic from AIs is still small in volume: between 1% and 3% of the total on a typical site. But the quality is in another league:
Those who arrive from an AI do not come to compare: they come determined. The AI has already recommended your brand before the click. That’s why 73% convert on their first visit, compared to 23% of organic traffic from Google.
With that level of intent, not measuring this channel is giving away the only proof that your strategy GEO is generating business.
How to measure traffic from AIs in Google Analytics: the 5 steps
The solution is a custom channel group. It is set up once and takes about 30 minutes.
- Step 1: go to Admin → View Settings → Channel Groups.
- Step 2: duplicate the default group and name it «Default + AI Traffic». Never edit the original.
- Step 3: create a new channel called «AI Traffic», with the condition: Source matches the regular expression (that’s what the option is called; the list to paste is below).
- Step 4: drag the new channel above «Referral». Google Analytics evaluates the rules from top to bottom: if Referral is first, it will keep your AI visits.
- Step 5: save and select this group in your Acquisition reports.
The list of AI domains for step 3, to copy and paste as is:
^(chatgpt\.com|chat\.openai\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|meta\.ai|deepseek\.com|grok\.com|x\.ai|poe\.com|you\.com|mistral\.ai)$Two details that prevent headaches:
The list is case-sensitive. Keep it all in lowercase, as it is here above.
The channel group does not work retroactively. Start classifying from the day you create it. That’s why it’s advisable to do it today, even if the volume is still small.
Do you want to see the historical data anyway? Go to Reports → Acquisition → Traffic Acquisition, and filter the dimension «Session Source» with «chatgpt» (or the AI you want to check). That does work with old data: the visits were always there, just without their own channel.
Which AI sends you the most traffic?
Once the channel is set up, you will see something like this:
ChatGPT domina con el 87% del tráfico de referencia desde IAs. Después vienen Perplexity (4%), Claude y Copilot (2% cada uno).
And Gemini? It still has a small impact, but it’s the fastest-growing source: +388% year-over-year. With Google pushing AI across all its products, it’s worth having it on the list starting today.
The «dark traffic»: what Google Analytics will never show you
Here goes the warning that is missing in most guides.
Even if you configure everything perfectly, you will only see between 30% and 40% of the visits influenced by AI.
The rest is hidden in three places:
- Direct: the person copies and pastes the link, or the mobile app opens the browser without passing the source.
- Organic search: the AI mentions your brand without a link, and the person later searches for you on Google by name.
- Generic referral: everything that your list does not capture.
There is no configuration that can fix this.
What you can do: add the question «How did you hear about us?» in your forms, with «an AI (ChatGPT, Gemini…)» as an option. It’s the simplest way to measure the invisible.
The rule for reading your reports: what you see in Google Analytics is the floor, not the ceiling. The real influence of AIs on your visits is 2 to 3 times greater than what the channel shows.
The special case: Google AI Overviews
Clicks from AI Overviews do not appear as AI traffic: Google counts them as normal organic search, without distinguishing them.
Today there is no clean way to separate them in Google Analytics. There is an advanced trick with Google Tag Manager (detecting links with highlighted text that Google uses), but it is imprecise.
The practical: treat AI Overviews as part of your organic and measure there what you can control, which is whether you appear cited or not.
The most common mistakes (and how to avoid them)
| Error | Consequence | Solution |
|---|---|---|
| Leave the AI channel below Referral | Referral takes all the AI visits | Drag «AI Traffic» above Referral |
| List in uppercase | Matches nothing and the channel remains empty | Everything in lowercase |
| Track only ChatGPT | You miss the growth of Gemini and Perplexity | Use the complete list and review it from time to time |
| Wait for historical data | The channel group only ranks since its creation | Create it today and use the source filter for the past |
| Trust only the AI Assistant channel | You see all AIs mixed, with no history prior to May 2026 | Create your own channel and filter by Session Source |
| Read Google Analytics like the complete picture | You underestimate the real influence of AIs 2-3 times | Add the survey «How did you find us?» |
Google Analytics measures after the click. What about before?
Everything you configured so far measures one thing: what happens after someone clicks.
But most of the influence of AIs happens before: in the response itself, when ChatGPT or Gemini decide whether to mention you or recommend your competitor.
That half is not visible in Google Analytics. It is seen by monitoring the responses.
That’s what we do at CreceRank: you set up the prompts for your industry and the platform queries them every day in ChatGPT, Gemini, Perplexity, and Google AI Overviews.
There you see if you appear, in what position, and against whom.
Google Analytics tells you how many arrived. CreceRank tells you how many were talked about you.
→ Start 14 days free, no credit card required
Conclusion
So, how to measure the traffic that AIs send you in Google Analytics? In three moves:
- Create the «AI Traffic» channel group today with the list from this guide. It takes 30 minutes and does not classify backwards: every day that passes is data you lose.
- Read the numbers as a floor, not a ceiling: 60-70% of AI influence comes disguised as direct or organic traffic. The survey «How did you find us?» completes the picture.
- Mide también lo que pasa antes del clic: si las IAs no te mencionan, no hay tráfico que medir. Empieza por el verificador de visibilidad en IA.
Traffic from AIs is small, but it has the highest intent that exists today. Measure it before your competition understands why it benefits them to do so.
For the full market context, see our AI usage statistics in Spain and Latin America, updated with original data.
FAQs
Sources
- The Agency Guide to Tracking AI Traffic in Google Analytics (Swydo): the complete setup, regex, and conversion benchmarks and dark traffic.
- How to Track ChatGPT Referrals in Google Analytics (Rankshift): the details of the utm_source=chatgpt.com parameter and its limitations.
- Google Analytics Adds AI Assistant As Default Channel Group (Search Engine Journal): the announcement of the AI Assistant channel and how it works.
