An analytics setup that accurately tracks AI search sources combines referrer detection, server log analysis, and UTM tagging, because no single tool sees the whole picture. Analytics platforms catch visits with recognizable AI referrers, but mobile apps and in-app browsers strip referrer data, which is where the measuring referral traffic from ChatGPT gap comes from. OpenAI reports ChatGPT passing 500 million weekly users, and SparkToro's zero-click study found only 37% of US Google searches end in a click, so a meaningful share of that traffic never shows up in standard dashboards.
Why Do Analytics Tools Miss AI Referrals?
Three reasons. Some AI engines pass no referrer at all, so the visit appears as direct. Some pass a generic referrer that analytics cannot classify. And mobile app surfaces strip referrer data entirely as part of their privacy model. The result is that AI traffic undercounts, often by a wide margin, unless you add a second measurement layer.
The fix is log-based tracking. Your server sees every request regardless of referrer, so logs capture the visits analytics tools classify as direct. The how AI crawlers access your site logging setup gives you the same raw view for humans arriving from AI answers.
What Should Your Analytics Stack Include?
Use three layers together. First, your analytics platform with AI-referrer rules that classify known AI sources. Second, server log analysis that catches untagged AI visits and lets you match them to user agents and referrers. Third, UTM tagging on any AI distribution you control, so those sources are explicit in your dashboard.
The AI citation attribution model depends on this stack, because attribution is only as good as the source data underneath it.
How Do You Identify AI Traffic in Server Logs?
Filter logs for known AI referrer domains and for the device and session patterns AI answer traffic produces. Match referrer hostnames to engines like ChatGPT, Perplexity, Gemini, and Bing AI surfaces, and correlate request timing with known engine activity. Over time, you can build a classifier that flags AI-driven visits.
We use log analysis as the ground truth in our AEO/SEO monitoring, because it does not rely on what the analytics vendor chooses to classify.
What About In-App and Zero-Referrer Traffic?
Zero-referrer traffic is the hard case. Without a referrer, the visit looks direct, and no dashboard can tell you its true source. Log analysis helps by showing which pages get spikes in direct traffic, which correlates with citations, and by matching request patterns. It is an inference, but a reliable one when citation and referral data line up.
For a cleaner signal, track citations alongside traffic: a cited page that also shows a direct-traffic spike is almost certainly drawing from AI answers. The two signals confirm each other.
How Do You Turn Tracking Into Decisions?
Route the measured AI traffic into your conversion funnel and compare it against other channels. Once you know which cited pages drive sessions and which of those convert, you can spend effort on the pages and engines that produce revenue. The B2B GEO pipeline work makes the connection explicit.
The AI referral funnel setup then optimizes the landing experience for the traffic your tracking now reveals.
How Conbersa Tracks AI Search Sources for Clients
Conbersa runs a three-layer tracking setup for clients as part of its managed AEO/SEO service: analytics referrer rules, server log analysis as ground truth, and UTM-tagged AI distribution. Citation data, referral traffic, and conversion data are joined in one view so clients see the real impact of AI visibility.
We built this because undercounting hides the value of GEO. When AI traffic shows up as direct, the channel looks weak and gets defunded. Give the traffic a source, route it to the funnel, and the investment becomes obvious.