Somewhere in your Google Analytics dashboard, a cluster of visitors is behaving oddly. They land on a single page, often deep in your site, stay for a surprisingly long time, and convert at a rate your paid campaigns would envy. Then they vanish into the catch-all bucket labelled “direct” or “unassigned”, leaving no clue as to where they came from. A growing number of these visitors did not stumble onto your site at all. They were sent there by ChatGPT, Perplexity, Gemini or Google’s AI Overviews, after your content was cited as the answer to a question.
For a business that has invested in AI search visibility, seeing this clearly is important. If the traffic a citation generates cannot be seen, it cannot be reported on, defended in a budget meeting, or used to decide which content to produce next. Isolating AI referral traffic in GA4 is one of the more practical, underused steps in an AI search strategy, and it takes a modest amount of setup rather than a specialist platform.
Why This Traffic Hides From Standard Reports
GA4’s default channel groupings were built for a web of links, not a web of conversational answers. When a user clicks a citation inside ChatGPT or asks Perplexity a question and follows the source link, the referring information that reaches your analytics is often inconsistent. Some platforms pass a clean referrer domain, some strip it entirely, and some route the click through an intermediate redirect GA4 cannot interpret. The result is that AI-driven sessions scatter across “Direct”, “Unassigned” and occasionally “Referral”, collecting nowhere a marketing team can study them as a single group.
This is not a reason to ignore the data. It is a reason to build a small amount of custom reporting around it, much as analysts once taught GA4 to recognise paid social or email campaigns before those channels had proper defaults.
Setting Up GA4 to See the Signal
The most reliable starting point is a custom channel group or exploration segment built on referrer and session source matching. In practice, this comes down to four steps.
- Build a regular expression matching the domains of the AI platforms relevant to your audience, such as chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com, and apply it as a custom channel group condition in GA4’s Admin settings.
- Create a saved exploration that filters sessions by that custom channel, so the data sits one click away instead of buried inside a blended report.
- Cross-check landing pages against content you already know has been cited by AI platforms, using server log data or a citation-tracking tool where one is available.
- Review the “Unassigned” and “Direct” channels periodically for landing pages that make no sense as direct type-ins, such as a long-tail blog post nobody would type from memory.
None of this requires a new analytics platform. It requires treating AI referral traffic as a reporting problem to be solved, the way referral spam and dark social traffic were solved in earlier years.
What the Data Tends to Show Once You Look
The pattern that shows up most consistently once businesses isolate this traffic is a sharp gap between volume and value. Ahrefs’ own analytics data found that AI search visitors made up roughly 0.5 per cent of its total traffic, yet accounted for 12.1 per cent of signups, a conversion rate around 23 times higher than organic search. Separate industry analysis of web traffic in March 2026 found ChatGPT alone driving around 2.7 million visits, with Claude’s referral traffic growing the fastest month on month even though its absolute volume remains smaller than ChatGPT’s or Google’s.
The explanation is intuitive once you sit with it. Someone who asks an AI platform a detailed question and clicks through to a cited source has usually already received an answer. They are verifying a specific claim, checking pricing, or confirming a business genuinely offers what the AI said it does, not browsing idly between options. That makes for a warmer visitor than most organic search traffic, and the conversion numbers tend to confirm it.
Connecting the Dashboard to Decisions
None of this is useful sitting in isolation. Once AI referral sessions are visible in GA4, they become a second lens alongside share of voice as a KPI for AI search performance: share of voice shows how often you are cited, while referral tracking shows what happens after someone clicks through. A page cited frequently but converting poorly once visitors arrive is a different problem to a page rarely cited at all, and each calls for a different fix. That distinction earns its place in planning conversations, since a content team deciding which pages to refresh next, or a finance team weighing how much budget an AI search programme deserves, has a stronger case when it can point to session and conversion data instead of citation counts alone. Teams also tend to trust a GA4 export more readily than a third-party visibility score, simply because it sits inside the same tool already used to judge every other channel.
A Practical Starting Checklist
Before treating AI referral tracking as solved, it helps to confirm a few basics are genuinely in place.
- A custom channel group or saved segment exists for the main AI platforms your customers are likely to use.
- Landing pages inside that segment have been spot-checked against known cited content.
- Conversion events are mapped consistently, so AI referral sessions compare fairly against organic and paid channels.
- Someone on the team reviews the “Unassigned” bucket at least monthly for AI traffic that has slipped through.
- The resulting figures reach whoever owns content planning, not just a marketing dashboard nobody else opens.
None of these steps demand new tooling or a large budget. They demand a willingness to stop treating “Direct” as a dead end and start treating it as a channel with a story still to tell.
Conclusion
AI search is already sending visitors to Singapore businesses that have earned a citation, and the businesses that can see this traffic clearly are best placed to act on it. A few hours spent building the right segments in GA4 turns a vague sense that AI search is working into a dataset that can guide content decisions and justify budget with evidence. If you would like a hand setting this up properly for your own site, feel free to get in touch with the team at AISEOAgency.