Every day, people are realizing that AI chatbots are a great way to get information. Some of these people are in your target audience.
For a lot of these searches (or prompts?) the user has no interest in visiting a website. They just want a quick answer.
Other times, they’re looking for detailed information and want to read an article. Or they’re wondering where the answer came from, and they click on the source. Or they really need help and want to see a website.
For some searches, the visitor wants to visit a website.
For others, they do not.
This chart shows when various types of searches (and related content) became “no click” searches:
When the visitor does want to visit a website, and when they click through from an AI chatbot, you can measure that traffic in GA4.
In this post, we show how to track website traffic from AI using GA4. We’ll use three methods:
- GA4 filter (simple, fast, some data)
- GA4 exploration (detailed, more data)
- GA4 custom channel group (a bit more complex, but useful everywhere!)
It’s kinda like small, medium and large. The first one is just a basic GA4 skill. The other two are build-once-use-forever reporting methods.
Caution: Some AI users are on apps, not websites, and traffic from apps is categorized as “Direct traffic” in GA4. So it’s not possible to accurately report all AI traffic separately. That’s a bummer but GA4 isn’t totally accurate anyway. Traffic from all sources is underreported.
1. Track AI traffic with a GA4 filter
First, the simple way!
Add a filter to your pages report so the session source is “chatgpt.com” and you’ll see traffic from the most popular AI tool. Here’s what the GA4 report for AI traffic from ChatGPT will look like:
Here are the steps:
- Go to the Engagement > Page path and query string report.
- Set a long date range
- Click the “Add filter +” button at the top of the report
- In the filter settings on the right, select “Page Referrer” as the dimension. Select “exactly matches” as the match type. Enter “https://chatgpt.com” as the value
- Click “Apply”
That’s it. Don’t be surprised if there isn’t much there yet. This is a new channel. If there is traffic, probably the trend line shows that most of the traffic is very recent. In the screenshot above, you can see that there was almost no traffic from ChatGPT at all and then… a steady stream.
No traffic from AI at all? Check out our 8 ways to get AI to recommend your brand.
If you do see traffic, look closely at the report. What pages are attracting traffic from AI? Do you see any conversions yet? Are those visitors more or less engaged than visitors from traditional organic sources like Google Search?
Of course, there are other AI traffic sources beyond ChatGPT. Next we’ll make a report that captures much more data.
2. Track AI traffic with a GA4 exploration
Next we’ll do something a bit fancier. We’ll report AI referral traffic with a GA4 exploration. Explorations are custom reports that are saved in the “Explore” section, so you can go back to them again and again. They’re also very flexible (and fun) to make.
Here’s a preview what the final GA4 exploration will look like:
Here are the steps for making a GA4 exploration that tracks traffic from AI sources:
- Click on the Explore icon on the far left, then click “Blank” to create a new exploration
- In the top left, give your exploration a name, such as “Traffic from AI platforms”
- Set a long date range, such as “Last 12 months”
Now we’ll make a segment to show only traffic from AI sources…
- Click the plus sign (+) next to SEGMENTS
- Click the “Create a new segment” button and click “Session segment” then give the segment a name such as “AI traffic”
- Click “Add new condition” and search for then select “Page referrer”
- Click “+ Add filter” and set the condition to “matches regex”
- Copy and paste the following regular expression into the box and click “Apply”
^https:\/\/(www\.meta\.ai|www\.perplexity\.ai|chat\.openai\.com|claude\.ai|chat\.mistral\.ai|gemini\.google\.com|chatgpt\.com|copilot\.microsoft\.com|copy\.ai)(\/.*)?$|.*\.ai.*|.*\.openai.*|.*\.groq.*|.*\.metaai.*|.*\.meta\.com/ai.*
 |
Jes Scholz, Organic Marketing Consultant“While traffic contributions are low, setting up GA4 reporting using a regex on page referrer sessions is vital to understanding in which chatbots your brand is visible. Because when these traffic sources are hockey sticks (not if, when) you need to have already put in the work on any surface where you brand was not well optimized.“ |
This regular expression simply describes the domains of all of the major, current generative AI platforms: ChatGPT, Perplexity, Gemini, Microsoft Copilot, Claude, etc. It does not show traffic from Google’s AI Overviews, which is still categorized as organic traffic. You could use this same regex in the filter we created above!
Here’s what the segment will look like:
Next we’ll add the dimensions and metrics…
- Click the plus sign (+) next to DIMENSIONS
- Search for then select “Page referrer” and “Landing page + query string” then click “Confirm”
- Click the plus sign (+) next to METRICS
- Search for then select these three metrics: “Sessions” “Engagement rate” and “Session key event rate” then click “Confirm”
Almost done! All the pieces are here. Now we build the report…
- Drop or select your segment in the SEGMENT COMPARISONS box
- Drop or select your dimensions in the ROWS boxes
- Drop or select your metrics in the VALUES boxes
I like to change my CELL TYPE to “Heat map” but that’s just my personal preference.
Look closely at the report. Which AI platforms are sending the most traffic? Which pages attract traffic from which platforms? Is there a “key event rate” showing yet?
Do you see lots of (not set) as the landing pages? There are several possible reasons for that. None of them are concerning. Here, GA4 expert, Dana DiTomaso explains:
 |
Dana DiTomaso, Kick Point“There are a few reasons why you have (not set) traffic in GA4 but the most common reason is that you have people who leave your website open for a long time. If someone leaves your website open long enough in their tabs that their GA4 session times out, GA4 would need a page refresh in order to start up a new session. But think of how you revisit a tab that you’ve had open for a while — you probably don’t refresh it before scrolling or clicking around, right? When that happens, GA4 can’t set the source, medium, or landing page for that new session, and that’s how you end up with (not set) traffic. It isn’t great and I wish GA4 would handle it differently but for now that’s what we’re stuck with.” |
3. Track AI traffic with a GA4 custom channel group.
There are a lot of ways to customize GA4. My friend Dana DiTomaso knows them all. We had a short conversation at a recent conference about AI as a traffic source. A few days later she sends an email with a great idea and some simple instructions. Here is her post on the topic. And here’s how it works…
In GA4, the “channel groups” are the ways that Analytics categorizes traffic sources. The defaults are good, but they don’t account for AI traffic. But you can create a new channel group that has an AI channel that has AI as a channel, then use this to see traffic from AI sources separately.
Tip: Creating and changing channel groups doesn’t change the underlying data in Google Analytics. If something isn’t working for you, you can just delete it. You’re not going to break anything.
First we’ll make a new channel group. Then we’ll make a new channel inside it.
- Click the gear in the bottom right to get to the Admin section. Then in the Data display box, click “Channel groups”
- Click the “Create new channel group” button. Name your group. I named mine “AI Sources Group”
- Click the “Add new channel” button. Name your channel. I named mine “AI Tools”
- For the channel conditions, set the “Source” to “matches regex” then copy and paste in the following regular expression. Again, this simply describes websites that are AI tools and potential sources of traffic.
^https:\/\/(www\.meta\.ai|www\.perplexity\.ai|chat\.openai\.com|claude\.ai|chat\.mistral\.ai|gemini\.google\.com|chatgpt\.com|copilot\.microsoft\.com|copy\.ai)(\/.*)?$|.*\.ai.*|.*\.openai.*|.*\.groq.*|.*\.metaai.*|.*\.meta\.com/ai.*
- Save channel
- Click “Reorder” and drag your new channel up, so it’s just above the Referral channel.
Note: The order is important! Google Analytics looks at the top of this list first. If the session doesn’t meet the criteria of the first channel, it goes to the next channel. Our goal is to intercept referral traffic from AI sources and categorize it separately, so we want Analytics to consider our new channel before categorizing the session as just another referral.
- Click “Apply” then click “Save Group”
Give it a few minutes, then check your Acquisition > Traffic acquisition report. Choose a nice long date range then from the dropdown above the first column, select your new channel group as the primary dimension.
The report will look something like this:
I highlighted the traffic and the conversions to show the number of visits and leads generated from AI sources.
If the numbers look low, plot the row from these “AI tools” and look at the trendline above. Going up? Remember, this is just the beginning…
The future of content discovery
Until recently, search engines and social media were really the only channels for content discovery. And now, AI is a new channel for content discovery. It’s a new part of your SEO strategy. Eventually, your favorite SEO tool will report on traffic from AI search.
Yes, it’s new. But if you’ve been in digital for a while, you’re used to disruption. Remember when social media was new? All the same questions were asked. What will work? Who owns it? How will results be measured?
We’ll ask a few of those key questions here, with the hope that the data from the above reports can provide some answers:
- What content attracts traffic from AI sources?
- Visitors from which AI platforms are the most engaged?
- Visitors from which AI searches are most likely to convert?
- How can we make our brand more visible in this channel?
To answer that last question and prepare for this new area of digital marketing, we’ve put together this video. It shows how to get AI to recommend your content and your brand:
Finally, let’s remember that search engines, social media and now AI chatbots are great, but there are other ways to connect with an audience. Some of those have nothing to do with algorithms. They cut out the big tech middleman. They are hard to disrupt.
These channels are all about humans, relationships, fans and friends…
