Most e-commerce brands have no idea how many visitors they are getting from AI search engines. They assume the number is small, shrug, and move on. That assumption is increasingly wrong. ChatGPT's Browse feature, Perplexity's answer engine, and Gemini's search integration are all sending real referral traffic to websites right now, and most of that traffic is invisible inside default Google Analytics 4 setups.
This guide walks through exactly how to find it, segment it, and make sense of it, so you can finally see whether your AI visibility efforts are translating into actual visitors and conversions.
Why AI Referral Traffic Is Hard to Spot by Default
Google Analytics 4 does not automatically categorise traffic from ChatGPT or Perplexity as anything meaningful. Depending on how the visitor arrives, you might see it lumped into "Referral", "Direct", or "(not set)". Sometimes it shows up correctly with the referring domain visible, sometimes it does not.
There are three reasons for this:
- Referrer stripping: Some AI engines open links in a way that strips the HTTP referrer header entirely, so GA4 sees the session as direct traffic.
- No UTM parameters: Unlike paid ad platforms, ChatGPT and Perplexity do not append UTM tags to the links they generate. You have to build your own detection logic.
- Inconsistent behaviour across features: ChatGPT's web browsing behaves differently from its citation links in the sidebar. Perplexity's "Sources" links behave differently from its "Related" section. The referral data is inconsistent even within the same platform.
The result is that most brands are dramatically under-counting AI-driven traffic. Fixing this properly requires a combination of referral source identification, custom channel groupings, and, ideally, some server-side logging.
Step One: Check Your Referral Report First
Before setting anything up, go and look at what you already have. In GA4, navigate to Reports > Acquisition > Traffic Acquisition. Set the primary dimension to "Session source/medium". Then search for the following domains:
chatgpt.comchat.openai.comperplexity.aigemini.google.comcopilot.microsoft.comyou.com
Pull a 90-day window. If you see any sessions from these sources, that is AI referral traffic that GA4 is already catching via the referrer header. Note the volume, bounce rate, session duration, and conversion rate. AI-referred visitors often behave very differently from organic search visitors, typically arriving with a more specific intent and converting at a higher rate.
If you see zero traffic from any of these domains, that does not necessarily mean you are getting none. It may mean the referrer is being stripped.
Step Two: Create a Custom Channel Group for AI Traffic
GA4's default channel groupings have no category for AI search. You need to create one manually. This lets you track AI referral traffic as its own segment over time, separate from organic, direct, and referral.
How to set it up
- Go to Admin > Data Display > Channel Groups in GA4.
- Click "Create new channel group".
- Add a new channel and name it "AI Search".
- Set the condition: Session source contains
chatgpt.comORchat.openai.comORperplexity.aiORgemini.google.comORcopilot.microsoft.com. - Save and apply the group.
GA4 will not backfill historical data with the new channel group, so the sooner you set this up, the sooner you start accumulating a clean dataset. Do this today, not next month.
Once active, you can use this channel group in the Traffic Acquisition report to see AI Search as its own row alongside Organic Search, Direct, Paid Search, and so on.
Step Three: Use UTM Parameters on Your Own Content
You cannot control the links ChatGPT or Perplexity generate when they cite your site. But you can control links you place in environments where AI engines might pick them up: press releases, structured data, knowledge panels, FAQ content, and anywhere else you publish URLs publicly.
If you add UTM parameters to these URLs before an AI engine indexes and surfaces them, there is a chance those parameters will survive into the referral. This is not guaranteed, but it costs nothing to try and occasionally it works.
A simple structure to use:
utm_source=chatgptutm_medium=ai-referralutm_campaign=ai-visibility
More practically, use UTM parameters on any link you share in a context where you have control: your Google Business Profile, your schema markup's url fields, syndicated content, and backlinks from sites you manage. This gives you a cleaner paper trail even if the AI strips the referrer.
Step Four: Set Up a Referral Exclusion (Carefully)
GA4 has a referral exclusion list under Admin > Data Streams > Configure Tag Settings > List unwanted referrals. This is normally used to prevent payment gateways from creating false new sessions. Do not add AI domains to this list. If you exclude perplexity.ai as a referral, GA4 will absorb those sessions into whatever the previous session source was, completely hiding the AI traffic. Leave AI domains off the exclusion list entirely.
Step Five: Supplement with Server Logs
GA4 only sees what the browser reports. If a user has tracking protection enabled, or if the referrer header is stripped at the network level, GA4 misses the session entirely. Server logs do not have this problem. They record every HTTP request, including the referrer header and the user agent string.
Check your server logs (or your CDN's access logs if you use Cloudflare, Fastly, or similar) for user agents associated with AI crawlers and browsers. You are looking for strings like:
ChatGPT-UserPerplexityBotGooglebot(for Gemini's indexing)OAI-SearchBot
Note that crawler visits and user referral visits are different things. A crawler hit means an AI engine is reading your content. A referral visit means a human clicked a link from an AI answer and arrived on your site. Both are worth tracking, but they tell you different things. Crawler activity tells you whether you are being indexed by AI engines. Referral traffic tells you whether that indexing is generating actual visitors.
If you want a deeper view of how AI engines are engaging with your site before the traffic even arrives, this server-log approach is worth combining with a proper AI visibility audit. At FlinnSchema's free AI visibility audit, we look at exactly this: whether your structured data makes you citable, and whether AI engines are actually crawling and using your content.
Step Six: Build a Simple AI Traffic Dashboard in GA4
Once your custom channel group is live and you have had a few weeks of data, build a dedicated exploration report so you can monitor AI traffic without hunting through the standard reports each time.
What to include in the exploration
- Dimensions: Session source, Session medium, Landing page, Country
- Metrics: Sessions, Engaged sessions, Engagement rate, Conversions, Revenue (if applicable)
- Filter: Channel group exactly matches "AI Search"
Run this as a free-form exploration with a date comparison so you can see month-on-month changes. As AI search referrals grow, you want a dashboard that makes the trend obvious rather than something you have to calculate manually each time.
Pay particular attention to landing pages. Which pages on your site are being cited by AI engines? Those pages are your highest-performing AI assets right now. They deserve fresh structured data, regular updates, and clear entity signals. If a product page is driving AI referrals, that product's schema markup should be spotless.
What Good AI Referral Traffic Actually Looks Like
When you start seeing real AI referral data, here is what tends to stand out. Sessions from Perplexity often have very high engagement rates, sometimes 70-80%, because the visitor has already had their question partially answered and is arriving to complete a purchase or read more detail. Sessions from ChatGPT's browsing feature tend to be more exploratory. Gemini referrals often look similar to organic search traffic in behaviour.
Bounce rates from AI referrals are frequently lower than from social media and comparable to or better than organic search. If yours are high, that is a signal worth investigating. It may mean the AI is citing your page for a query that does not match your actual content, which is a structured data problem rather than a traffic problem.
Understanding the quality of this traffic is just as important as knowing the volume. That is something we explore in detail in the post on which AI search engine sends real customers, which is worth reading alongside this one.
Connecting Traffic Data to AI Visibility Strategy
Tracking is only useful if it changes what you do. Once you have a working AI traffic setup in GA4, use it to answer three questions every month:
- Is AI referral traffic growing as a share of total traffic?
- Which pages are being cited, and do those pages have strong schema markup?
- Are AI-referred visitors converting, and at what rate compared to other channels?
If AI traffic is growing but conversions are low, the problem is usually on-page. If AI traffic is flat despite strong organic rankings, the problem is usually structural data and entity clarity. Both are fixable.
The brands seeing the most AI referral growth in 2025 and 2026 are not necessarily the ones with the most content. They are the ones whose content is structured in a way that AI engines can extract, verify, and confidently cite. Schema markup, clear entity definitions, and authoritative structured data all contribute to this. You can read more about how schema compares to raw content volume for AI visibility in our breakdown of the two approaches.
If you want someone to review your current setup and tell you exactly where AI traffic is leaking and what to fix first, start with a free audit. It takes about two minutes and gives you a concrete starting point rather than a generic checklist.
Frequently Asked Questions
Does GA4 automatically track ChatGPT and Perplexity as referral sources?
Sometimes, but not reliably. GA4 will capture sessions where the HTTP referrer header is intact and the source domain matches. However, many AI-referred sessions arrive without a referrer, meaning they are counted as direct traffic. You need a custom channel group and ideally server-log monitoring to get an accurate picture.
Can I use Google Search Console to track AI traffic?
No. Google Search Console only tracks traffic that arrives via Google Search properties. It will not show you clicks from ChatGPT, Perplexity, or other third-party AI engines. GA4 and your server logs are the right tools for this.
How do I know if my traffic data is accurate enough to make decisions from?
Look for consistency over at least 60 to 90 days. A single month of data can be noisy. Once you have two or three months with a custom channel group in place, you can start identifying real trends. Cross-reference against your server logs for pages that are seeing crawler activity but not generating referral sessions, as that gap is worth investigating.
What if I am getting zero AI referral traffic right now?
Zero traffic usually means one of three things: AI engines are not indexing your content, they are indexing it but not citing it in answers, or they are citing it but the referrer is being stripped and the visits look like direct traffic. The first two are fixable with better structured data and entity signals. The third requires server-log monitoring to detect. An AI visibility audit is the fastest way to diagnose which problem you have.