Analysing AI Traffic in GA4

The digital landscape is currently undergoing its most significant shift since the arrival of the smartphone. As generative AI becomes a primary interface for how people discover information, a new class of visitors is arriving on our websites: AI-referred user and AI agents. While these visitors may still only represent a small percentage of total volume, their impact is disproportionately high, often converting at rates three to four times higher than traditional search.

For businesses, the ability to surface, measure, and analyse this traffic in GA4 is no longer just a technical exercise; it is a strategic necessity. To ignore these signals is to fly blind in an era where AI doesn’t just answer questions, it directs the future of consumer behaviour. 

Defining the Scope

Before we dive into the data, it is important to establish the boundaries of what GA4 can actually see. This article is focused exclusively on the traffic that successfully reaches your website. It will not cover ‘Zero-Click’ activity, the instances where an AI provides an answer using your data without the user ever clicking through. Furthermore, because GA4 relies on browser-based signals, we cannot report on traffic from native AI apps (which often strip out all tracking data) or AI tools that do not pass a referrer header.

We are looking at the ‘visible’ portion of the AI iceberg: the users and agents that actually land on your pages and interact with your brand. 

Getting started

The best way to get started is to review what AI traffic you are already capturing in GA4. To do this look at a Traffic Acquisition report. Unlike Explorations, Reports are not limited by your retention period (max 14 months for standard properties), so you can run this for a much longer date range to ensure you capture most traffic sources, especially if your website isn’t receiving high traffic volumes overall. 

Once you’ve selected your date range in your Traffic Acquisition report, add “Session source/medium” as a secondary dimension and review the sources that you want to group as AI traffic, so you can write a regular expression (regEx) you can use as a filter. 

Depending on how thorough you are with your source review this can be very time consuming, so we’ve prepared a regEx for you that you can copy and paste to save you some time: 

.*\.ai.*|.*openai.*|.*copilot.*|.*chatgpt.*|.*gemini.*|.*perplexity.*|.*poe.*|.*claude.*|.*andisearch.*|.*grok.*|.*phind.*|.*edgepilot.* 

Copy this regEx and go to customise your report by applying a filter to the session source/medium dimension, for Match Type select “matches regex”, paste in the above and apply. Make sure your report still contains session source/medium as a secondary dimension. You can now review your AI traffic. 

You could also use this as an exclusion filter by selecting Match Type “does not match regex”. This way you can review your remaining sources to validate anything that was missed and should be added to your regEx. 

While the regEx we provided you with may not seem very extensive, it should cover the key AI traffic drivers. The list is deliberately kept short and simple as filters in the GA4 interface are limited to 250 characters. 

Getting organised

In theory you are ready to go now to analyse your AI traffic and performance but solely relying on filters or possibly segments isn’t the most convenient way to go about when you want to delve deeper into your data. 

We recommend setting up a new custom channel group for your AI Traffic. To do that go to the Admin section in GA4 and under Data Display select “Channel groups”. If you have already set up your own channel groups here you might choose to add this new AI Traffic channel to an existing one or if you don’t have any set up yet, you can create a new one now. 

Whichever option you choose (editing an existing one or creating a new one) you will have to add a new channel. Give your new channel a meaningful name such as AI Traffic and for the channel conditions select “Source” as a condition group, then add “matches regex” as the condition.  

If you are happy the short regEx you used as a filter earlier is sufficient to categorise your AI traffic you can use this to define your new channel. However, if you are concerned about missing anything you can write a much longer regEx for you channel group as the character limit here is far more generous but be careful that you don’t accidentally match against sources you don’t want to class as AI. 

Once you’ve saved your new channel it will go to the bottom of your channel list. As channel groups work on a hierarchy it is very important that you reorder your channel list because GA4 looks at this list in order, starting from the top, checking until a condition is met. As most of the uncategorised AI traffic currently falls under Referral or Organic Search you need to make sure that your new AI Traffic channel sits above those two channels by clicking the “Reorder” button and dragging the group up to where you want it. You can now save your channel group. 

As custom channel groups work retrospectively you can run a report or exploration against this new custom channel to validate all sources have been categorised correctly. To do this use a Traffic Acquisition report (or an Exploration) and select “Session Name of your Custom Channel Group” as a primary dimension and “Session source/medium” as a secondary dimension. As your main focus is to sense check traffic categorisation your key metric in this report will be sessions. Now you can see if your regEx needs redefining or if the order of your channel groups needs redefining. Make sure your AI Traffic doesn’t include sources that should not be classed as AI e.g. affiliate or paid search traffic or if there are still known AI tools that come under Referral instead of your new channel. 

You should probably review your custom channel groups on a regular basis to ensure sources are still categorised correctly. Once or twice a year should be sufficient unless you become aware of a surge in new AI tools you want to include or if you are actively making changes to your website to make it more accessible for AI. In those cases, you might want to review this more often to ensure you categorise all traffic that’s important to you correctly. 

A note on Primary Channel Groups:

If you have edited an existing custom channel group that was set to be your Primary Channel Group or if you made your new custom channel group your Primary Channel Group, be aware that this will only work moving forward, despite custom channel groups in general working retrospectively. While we would recommend setting the custom channel grouping that contains your AI Traffic channel as your Primary Channel Group you need to be mindful that this will affect some of your reports, saved explorations and automated reports that rely on this dimension. If you decide to go ahead with this, we’d advise to add an annotation to GA4 that states when this new Primary Channel Group has come into effect or when the change to an existing one was made. 

Gaining Insights

Now that you’re finally set-up, the fun part begins: diving deeper into the data and gaining insights. 

This is also where the possibilities of what to look for are (almost) endless, so the below should give you some ideas of what you could do with this information, although ultimately it entirely depends on your specific goals and what is important to you. 

Traffic and Actions:

When you set up and validated your filters and/or custom channel you already looked at sessions, so by now you will have a pretty good idea of traffic volumes to your site and what AI tools are driving this traffic.  

To review the quality of this traffic you could look at what actions or events this results in. In your reports or explorations you can bring in event names to see what actions were completed e.g. events like add_to_cart or purchase are probably valuable actions you want to focus on. 

Engagement:

If you add in engagement metrics like Bounce Rate, Engagement Rate, Average Session Duration or Average Engagement Time per Session you may notice that users arriving on your website from AI sources are highly engaged and achieve a better conversion rate than users coming from other sources. This is most likely because they already received the info they were looking for via the AI overview or chatbot and clicked through on a link ready to take action as they already know what to expect when they land on your site. 

High engagement, zero bounces and long session durations far above the average (human) user behaviour could also be an indicator of AI Agents deep scanning your site. AI bots don’t get distracted and thoroughly parse your content, compare prices, summarise blog posts or complete actions like filling in forms, adding items to basket or even making purchases. 

High engagement from AI traffic is often a sign that your site is “Machine Readable.” This is a huge win for SEO in 2026, as it means AI search engines find your content valuable enough to spend time “digesting” it for their users. 

Funnels and Paths:

Although engagement is high and some actions are completed you may want to review if there are any technical stumbling blocks for AI agents to complete certain tasks such as making it all the way through your checkout process. 

You can use the predefined Funnel exploration in GA4 and add a segment for your AI Traffic either based on your newly created session custom channel or using the regEx to match the session source. Review the predefined steps and adjust them to match your checkout process to identify barriers based on high abandonment rates. 

You could also look at a Path exploration applying the same segment for your AI Traffic and then select to “start again” in the visualisation panel of your exploration. Start from the end point by defining a relevant event e.g. purchase or page e.g. confirmation page and work your way back through the steps to see how AI agents or users coming from AI sources navigated through your site to reach the specific end point. 

Landing Pages:

To understand which landing pages AI bots most interact with you could create an exploration that looks at landing pages to see how many sessions they receive, what the bounce rates and engagement rates are and so on.  

Once you’ve identified your top pages that receive highly engaged AI traffic, your “AI Magnets”, you could review what structure those pages have in common and apply this to the rest of your pages. 

You could go even more granular and review landing pages by session source in case certain AI agents drive more valuable actions (e.g. sales) than others and optimise towards those. 

Role in the Journey:

Last but not least it might be interesting to see what role AI traffic plays in a purchase journey. Do they initiate the journey and bring in new customers? Do they assist during the decision-making phase or do they close the deal? 

For this you could take a look at Attribution Paths under the Advertising section in GA4. Unfortunately, the only choices for your primary (and only) dimension here are Primary Channel Group, Default Channel Group, Source, Medium or Campaign. This means that unless you made the custom channel grouping that includes AI Traffic your Primary Channel Grouping you are unable to use this as a dimension here. Remember, if you did make it your Primary Channel Grouping it will only work moving forward, not retrospectively.  

(While you could look at Source, you can’t filter by this dimension and the report it produces at this level is probably too granular for you to see results at a glance.) 

If you have decided to make the custom channel grouping that includes AI Traffic your Primary Channel Grouping and some time has passed since then for enough key events to have happened, this report can show you whether AI traffic was an early, mid or late touchpoint. 

With this information you can see where your site is already doing well in terms of being machine readable and where you should focus your effort on making improvements. You should consider questions like: 

• Is there some content that should be more present in AI overviews during the discovery phase to top up the funnel with new potential customers OR are you already present and your content is so good that users get all the answers they want directly in the AI overview without having to visit your site? 

• Is this a competitive space and your content during a user’s research phase needs to be better structured to make it more machine readable? 

• Are there technical barriers in your checkout process that prevent AI agents from completing a purchase? 

Final Thoughts

As we move deeper into this AI-augmented era, the line between human browsing and automated agency will continue to blur. Success no longer depends solely on how well you appeal to a human eye, but how effectively your data can be “digested” and recommended by an algorithm.  

By implementing these GA4 configurations today, you aren’t just cleaning up your reports; you are building a feedback loop that allows you to optimise for the most influential gatekeepers of the modern web.  

The “visible” portion of the AI iceberg is growing, and those who take the time to measure it now will be the ones best positioned to navigate the waters of the future. 

Find Out More

While measuring this traffic in GA4 is the vital first step, the next challenge is learning how to influence which of your products these AI agents actually choose to surface. To move from observing the ‘iceberg’ to controlling your visibility within it, join our upcoming webinar series on Owning AI Shopping Journeys. 

Get In Touch

For more information on how we can help you when it comes to all things data, get in touch with the team today.