How to Track AI Search Referrals in GA4

AI assistants such as ChatGPT, Perplexity, Gemini, Claude and Copilot send visitors to websites when they cite sources or recommend tools. In GA4, those visits usually show up as ordinary referral traffic, mixed in with links from partner sites, directories and forums, which makes them hard to analyze. If you want to track AI search referrals in GA4 properly, a few configuration changes let you separate this traffic into its own channel and see how it converts.

This guide covers where AI referral traffic comes from, the limits of what analytics can see, and a step-by-step setup that a small marketing team can complete in an afternoon. It ends with how to combine the data with manual checks so you understand both how much traffic AI tools send and why they send it.

Where AI traffic comes from

When someone clicks a link inside an AI answer, the visit often arrives with a referring domain that identifies the tool. Common ones include chatgpt.com and the older chat.openai.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com. Smaller AI search tools appear as well, and new ones launch regularly.

These domains are not fixed. Tools rebrand, move to new domains and add apps that behave differently from their web versions. Any setup you build should be treated as a list you maintain rather than a configuration you finish once.

The limitations to understand first

AI referral data in GA4 is incomplete, and it is worth understanding why before you start drawing conclusions from it. Some AI apps do not pass referrer information at all, so their visits look like direct traffic. Mobile and desktop apps are especially likely to strip referrers. Some users copy a URL from an answer and paste it into their browser instead of clicking, which also records as direct. And much of the influence AI tools have is indirect: a buyer reads an answer that mentions your product, then searches for your brand later, and GA4 credits that visit to organic or direct search.

The practical consequence is that AI referral numbers are a floor, not a full count. They show the minimum traffic AI tools send and are useful for spotting trends, but they understate total impact. Keep that in mind when sharing numbers with colleagues, so nobody concludes that AI search does not matter because the referral line looks small.

Step 1: Check your existing referral data

Before configuring anything, look at what is already there. In GA4, go to Reports, then Acquisition, then Traffic acquisition. Change the primary dimension to Session source, and use the search box above the table to look for AI domains such as chatgpt or perplexity.

This tells you how much AI traffic you are already receiving and which tools send it. It also reveals any sources you might not have expected, which you will want to include in the channel definition in the next step. If you see almost nothing, that does not necessarily mean AI tools ignore you. It may simply mean the traffic is small so far, or arriving without referrers.

Step 2: Create a custom channel group

GA4's default channel grouping lumps AI visits into Referral. A custom channel group lets you give them their own channel, which then appears in standard reports and explorations.

Go to Admin, then Data display, then Channel groups. Create a new channel group, and the simplest approach is to copy the default group so all existing channels stay intact. Add a new channel and name it something clear, such as "AI Assistants." Set its condition so that Session source matches a regular expression covering the AI domains you care about. A starting pattern looks like this:

chatgpt.com|chat.openai.com|perplexity.ai|gemini.google.com|claude.ai|copilot.microsoft.com

Then move the AI Assistants channel above Referral in the channel order. GA4 evaluates channels from top to bottom and assigns each session to the first one that matches. If Referral sits above your AI channel, AI visits will be classified as referrals before your rule ever applies. Save the group.

Custom channel groups apply to historical data as well as new data, so once it is saved you can look back at previous months with the new channel in place. You can now select the group in acquisition reports and use it as a dimension in explorations.

Step 3: Build an exploration

Standard reports are useful for a quick look, but an exploration gives you more control. Create a free-form exploration and add session source, landing page and your custom channel group as dimensions. Add sessions, engaged sessions, key events and session conversion rate as metrics. Then filter to the AI Assistants channel.

The resulting table shows which pages AI tools send people to and how those visitors behave once they arrive. You will often find that a handful of pages receive most of the AI traffic. Those pages are telling you which questions AI users ask that your site answers well.

Step 4: Track conversions

Traffic numbers mean little without outcomes. Make sure your important actions, such as signups, demo requests and trial starts, are configured as key events in GA4. If they are not, set them up before drawing conclusions about AI traffic.

With key events in place, compare conversion rates from the AI channel with those from organic search. Look at which landing pages produce AI conversions, and compare engagement measures such as engaged session rate. AI referral visitors can have different intent from search visitors. Some arrive after a long conversation in which they have already narrowed their options, which can make them more ready to act. Others click a citation out of curiosity and leave quickly. Your own data will show which pattern dominates for your site.

Step 5: Add self-reported attribution

Because analytics misses much of AI's influence, ask people directly. Add an option such as "AI assistant (ChatGPT, Perplexity, etc.)" to the "How did you hear about us?" field on your signup or demo form. If you do not have that field, it is worth adding as an optional question.

Self-reported answers capture influence that GA4 cannot see, including the buyer who read about you in an AI answer and then typed your URL. The answers are imperfect, since people misremember or pick the first plausible option, but they are a useful counterweight to referral data that systematically undercounts.

Step 6: Monitor the landing pages AI tools link to

The pages that AI tools cite most are a strategic signal. They likely answer questions your buyers ask assistants, and they are your current foothold in AI search. Keep them accurate and up to date, since outdated content is a reason for any system to prefer a different source. Look at their conversion paths, because a visitor arriving from a citation needs an obvious next step. And consider creating related content that answers adjacent questions, building on the topics where you already have visibility.

Step 7: Watch trends over time

Review AI referral data monthly. Look at AI sessions, their share of total traffic, the conversions they produce and any new sources that have appeared. Single-month numbers bounce around, especially when volumes are small, so the trend over several months is more meaningful than any one month. A steady rise in AI sessions and AI-sourced conversions suggests your visibility is growing even if the absolute numbers remain modest.

Using Looker Studio to share the data

If others on your team need to see this regularly, a simple Looker Studio dashboard saves repeated report pulls. Connect it to your GA4 property and include AI sessions over time, top AI sources, top landing pages for AI traffic and conversions from the AI channel. Keep it small. A one-page view that answers "is AI traffic growing, where does it land and does it convert" is more useful than an elaborate dashboard nobody checks.

Combine analytics with manual testing

Analytics shows visits, but it cannot tell you why AI tools mention you or how they describe you. For that, test prompts directly. Make a short list of the questions your buyers are likely to ask, such as requests for tool recommendations in your category or how-to questions your content covers. Run them in the main AI tools periodically and note whether your brand appears, which pages are cited and how your product is described.

These checks often explain patterns in the data. If a page suddenly gets more AI traffic, a prompt test may show it has started being cited for a common question. If AI tools describe your product inaccurately, you know which pages or third-party sources need attention.

Keep your source list updated

New AI tools launch frequently, and existing ones change domains or add new surfaces. Review your referral sources each quarter, look for unfamiliar domains that might be AI tools, and update the regular expression in your channel group. A setup that was accurate six months ago can quietly miss a growing source if nobody maintains it.

A worked example (illustrative)

Imagine a small SaaS company selling scheduling software. Its marketer sets up an AI Assistants channel group and builds the exploration described above. In the first review, AI sessions are a small share of total traffic, and almost all of them land on two pages: a comparison of scheduling tools for clinics and a guide to reducing appointment no-shows.

Looking closer, the marketer notices that visitors from the comparison page convert to trials at a higher rate than organic visitors to the same page, while visitors to the no-shows guide mostly read and leave. Prompt tests explain the difference. Assistants cite the comparison page when users ask which scheduling tool to choose, so those visitors arrive ready to evaluate. The guide is cited for a general how-to question, so those visitors are earlier in their thinking.

The team responds in two ways. They update the comparison page with current pricing and a clearer trial call to action, since it is already doing commercial work. They add a short section to the no-shows guide that links to the relevant product feature, giving curious readers a natural next step. They also add an AI option to the signup form's attribution question. Over the following months, the monthly review tells them whether these changes are moving AI-sourced trials, even though the absolute numbers stay modest.

Mistakes to avoid

The most common setup error is leaving the AI channel below Referral, which quietly sends every AI visit to the wrong bucket. Another is writing the regular expression once and never revisiting it. A third is drawing conclusions from very small numbers in a single month. And the most consequential is treating referral counts as the full measure of AI influence, which leads teams to underinvest in a channel whose effects mostly show up elsewhere.

The takeaway

Tracking AI search referrals in GA4 takes a custom channel group, an exploration and properly configured key events. Treat the numbers as a minimum, supplement them with self-reported attribution and regular prompt testing, and review the source list every quarter. Together, those give a small team a reasonable picture of how AI assistants contribute to traffic and pipeline.