Referrers get stripped and buyers skip the click, so GA4 sees a fraction of what AI search does for you. ThriveStack citedby overlays server-log crawler hits, human traffic and revenue on one timeline, then ties citations to signups, deals and payments.
Works with GA4, your server logs, your CRM and your billing data.
AI attribution links an AI answer to the revenue that followed it. A buyer asks ChatGPT for a shortlist, reads your name, and shows up days later by a different route. Attribution connects those two moments.
A referrer alone cannot do that. ThriveStack citedby adds the signals that sit around it.









Because most of AI search happens before a click. The buyer reads the answer, forms a shortlist and leaves. Last-click reports see only what arrives with a referrer. Here is what the 2026 research shows.
26% of senior B2B marketing leaders believe AI shapes buying decisions without generating any clicks. Agentcy’s CTO Tom Fry put it plainly: “Buyers can now research, compare, and shortlist vendors without generating a single visit.”

Only 10% of leaders can consistently link AI touchpoints to revenue. 20% do not know what share of their website sessions come from AI tools.
In a 12-site GA4 test, total Direct sessions rose 140.9% from January to September while engaged Direct fell 1.0%. Direct sessions with identifiable AI traces averaged 0.003% of Direct.

Indig ran 632 tests of weekly AI visibility against Direct and branded search. The two promising correlations weakened once the test was made fairer. Correlation is not attribution.
Only ChatGPT and Gemini were seen adding utm_source in a scorecard of 11 AI engines, as of Oct 16, 2026. The ChatGPT tag also dropped off many links from Oct 9 to Oct 15, 2026. A tag you do not control is a weak foundation for attribution.
The experts disagree on the method. Two recent views sit side by side below. ThriveStack citedby brings in all the relevant signals (citations, crawler hits, AI visits, pipeline and revenue) to settle the debate with data.

Could not find hidden AI traffic in Direct. Suggests tracking brand mentions on a fixed prompt set, asking new customers where they first heard of you, and tying those answers to leads and revenue.

Says clicks measure the wrong thing. AI search works like a brand channel, effects lag, and referrers get stripped. Recommends holdout tests and adding AI share of voice to the marketing mix model.

Says precise attribution is breaking down because AI platforms do not report when they send traffic. Brand lift over time is the better measure.

Argues attribution models should weight brand visibility over clicks, since over half of Google searches now end without one.
They disagree on the method. They agree a referrer alone cannot carry the answer.
Sources: Growth Memo, iPullRank, Near Memo EP 244, Pulse.
The referrer is one signal. Server logs add the one most tools skip. AI crawler logs show every visit from an AI crawler, with or without a human behind it. ThriveStack citedby overlays three signals on one timeline.
You can then check whether crawler spikes, citations, visits and revenue move in the same order.
From your server logs. Which AI bots read which pages, and how often.
From GA4, split by channel. AI assistants get their own segment.
From your CRM and billing data, on the same days as the other two.
Traffic Attribution in ThriveStack citedby. Sample data.
AI Bot Activity counts crawler requests from your server logs.
Bot requests, human visits by channel and revenue share one chart, so you can see what moves together.
Three steps, from the first citation to the first payment.
Run AI search monitoring: track citations on a fixed prompt panel across 10+ AI engines. Read your server logs for AI crawler hits. Tag AI assistant visits in GA4.
Join visits to your CRM and billing data, so the AI origin stays on the record from the first visit to the first payment.
Report AI-sourced and AI-influenced pipeline and closed revenue. AI numbers are a floor, and the report says so.
The share of revenue AI earns, how that compares with its share of visits, and the revenue it influenced. Then the trend, so you can see whether each AI visit is worth more over time.
Revenue Attribution in ThriveStack citedby. Sample data.
In this sample, AI earns 30% of revenue from 13% of visits.
It adds buyers AI introduced who paid through another channel.
No single method settles it, so ThriveStack citedby supports four. They match what Indig and King each recommend.

Kevin Indig and Mike KingAI was the first place a buyer met you.
AI shaped the journey and another channel closed it.
Ask “how did you hear about us?” and join the answer to the deal.
Compare regions or product lines with and without AI work. Read results by quarter.
AI search attribution links an AI answer to the revenue that followed it. It should track four things: citations on a fixed prompt panel, AI crawler hits from server logs, human visits including AI assistants, and the signups, deals and payments that come after.
To track AI traffic in GA4, you rely on visits that arrive with a referrer or a UTM tag. Many AI visits arrive without one, and Direct traffic is full of bots, so GA4 alone undercounts AI and is hard to audit. Pair it with server logs and your CRM and billing data. See which engines send a UTM tag in our AI referral traffic scorecard.
Dark AI traffic is the influence of an AI answer that leaves no AI referrer behind. Experts disagree on how much of it sits in Direct. One 12-site GA4 test found almost none that was identifiable. Another view says the influence happens before any click. Measure it from several angles.
Server logs record every request to your site, including AI crawlers. Many of those crawlers do not run JavaScript, so a JavaScript analytics tag does not record them. Logs show which bots read which pages, and how often.
Experts split on this. Some say AI search works like a brand channel, so clicks undercount it and results lag by months. Others say clicks still matter. ThriveStack citedby reports both: click-level visits and revenue, plus influenced revenue from buyers AI introduced.
Many AI visibility tools stop at the mention: whether your brand appeared in an answer. ThriveStack citedby also reads your server logs, joins visits to your CRM and billing data, and reports AI-sourced and AI-influenced revenue.
It depends on your sales cycle. Brand effects can lag by months, so read the results by quarter. Use holdout tests by region or product line when you can.
Connect GA4, your server logs, your CRM and your billing data. See which AI citations turn into customers.