How to separate AI traffic from traditional organic traffic in GA4 — a technical guide
AI traffic is growing but hides inside "direct" in GA4. Here's how to create segments to capture AI referrers, measure that traffic's quality, and report the channel separately.
Traffic originating from AIs is the fastest-growing thing in digital marketing, but it's what most teams can least identify in their analytics. That isn't a GA4 failure — it's a structural problem with the channel: part of AI traffic arrives with an identifiable referrer (when the user clicked a link the AI generated), but most of it arrives as direct traffic (when the user read the AI's answer and then typed the URL or searched the brand directly).
This post focuses on the part you can capture: traffic with an identifiable AI referrer. It's a fraction of total AI traffic, but it's measurable, trackable, and representative enough for quality and trend analysis.
The AI referrers that reach your GA4
The AI platforms that generate traffic with an identifiable referrer mainly include:
perplexity.ai— the main generator of clicks with a referrer, because it cites sources with clickable links in every answerchatgpt.comandchat.openai.com— when the active-search version of ChatGPT links to sourcesclaude.ai— in smaller volume, when Claude with active search links outgemini.google.com— growing, especially for queries with result linkscopilot.microsoft.comandbing.com/chat
To check whether these referrers already reach your site, go to GA4: Reports → Acquisition → Traffic acquisition → filter by "Session source" and search for "perplexity," "chatgpt," and the other domains listed.
How to create an "AI traffic" segment in GA4
In GA4, create a comparative segment:
- Go to Explore → New exploration
- Under "Segments," click "+" and create a new session segment
- Name: "AI traffic"
- Condition: "Session source" contains any of:
perplexity.ai | chatgpt.com | chat.openai.com | claude.ai | gemini.google.com | copilot.microsoft.com | you.com - Save and apply to the exploration
With that segment, you can compare quality metrics between AI traffic and other channels: average session duration, pages per session, engagement rate, conversion rate for key events (sign-up, form fill, purchase).
How to configure GA4 to capture AI traffic more completely
Beyond the identifiable referrers, there are two configurations that expand coverage:
1. A custom channel-grouping definition. GA4 has a "Channel grouping" feature you can customize. Create a group called "Generative AI" and include the referrer rules listed above. That makes this traffic appear as a distinct category in the standard channel reports, not mixed into generic "Referral."
To configure: Admin → Property → Channel definitions → Custom channels → Add channel → Define referrer conditions.
2. UTM parameters on pages you control. If you have links to your site in your own materials (newsletter, presentations, directory profiles), adding ?utm_source=ai_referral&utm_medium=ai to those links doesn't track AI traffic directly, but it lets you segment traffic from users who came from contexts where AIs also reach them — creating an audience proxy.
The most important data point the AI segment reveals
The metric that most surprises teams when they create this segment for the first time is the conversion rate. Aggregate market data (Ahrefs, SparkToro, 2025 research) consistently shows that traffic originating from AIs has a conversion rate far above the organic-traffic average — in some studies, 9 to 15 times higher.
The explanation is behavioral: whoever clicks a link cited by an AI has already received a qualified recommendation. It isn't a chance discovery — it's a user who asked about a problem, received an answer that cited your company as the solution, and clicked to learn more. That level of pre-visit qualification is rare in other channels.
The AI traffic that still isn't capturable
To be honest about the limits: the larger part of AI traffic — the part where the user read the AI's recommendation and then searched your brand on Google or typed the URL directly — still has no standardized tracking solution. That traffic is in GA4 as "direct" or "branded organic" (searches for the company name), but you can't distinguish, with precision, how much came from AI influence.
Indirect methods for estimating that volume include: an increase in brand-search volume in Google Search Console correlated with AI-visibility actions, and qualitative research with new customers about how they discovered the company.
How Crowly complements GA4 in this analysis
Crowly measures exposure — how often your company is cited in the AIs. GA4 with AI-traffic configuration measures downstream conversion — how many of those users reach the site. Together, the two data points let you calculate a "citation-to-visit" rate that is the most useful channel metric currently available for AI.
Start by measuring exposure with Crowly's free diagnostic. Configure GA4 in parallel to capture the visits. Free diagnostic →
Sources:
