How to attribute revenue to the AI channel — the tracking problem that still has no perfect solution
Traffic from AI is growing 527% a year, but most companies can't measure it. Understand the AI channel's attribution problem and the solutions available today.
Traffic originating from AIs grew 527% in the first half of 2025 versus the same period the year before. It's the fastest-growing channel in digital marketing. And, paradoxically, it's the channel most companies can least measure — because traditional attribution models weren't built for it.
That measurement gap creates a real problem: marketing teams need to justify investment in AI-visibility strategies, but they can't show, with the precision a performance manager would demand, which revenue came specifically from that channel. That delays decisions, cuts budget, and, ultimately, leaves the company behind while competitors who tolerated the ambiguity have already built presence.
Why AI tracking is different from tracking other channels
In conventional digital marketing, the attribution model works like this: a user clicks an ad → the UTM in the URL records the source → analytics connects the click to the later conversion. The channel is tracked end to end.
The AI channel breaks that logic at two points:
Point 1 — The AI doesn't necessarily link to your site. When ChatGPT mentions your company in an answer, there's frequently no clickable link. The user read the recommendation, noted the name, and went to find you directly on Google or by typing the URL. That traffic arrives as "direct" in analytics — invisible as an AI origin.
Point 2 — When there is a link, the referrer can be lost. On AI platforms that generate links (Perplexity is the main one), the referrer perplexity.ai should appear in analytics. But depending on HTTPS settings, the type of click (direct in a mobile app vs. a browser), and the user's privacy settings, the referrer can be lost or masked.
Those two factors combined mean most AI traffic today, in most companies' analytics, is hidden inside direct traffic — the least informative category of all.
The AI referrers you can track today
Part of AI traffic is trackable via referrer. The domains you should monitor in GA4 as a specific "AI referral" segment:
chatgpt.comandchat.openai.comperplexity.aiclaude.aigemini.google.comandbard.google.comcopilot.microsoft.comandbing.com/chatyou.comphind.com
Create a GA4 segment with those referrers and track it weekly. The numbers will be smaller than the real AI traffic — because only the fraction that came with a trackable link will appear — but they're real data you can use to compare traffic quality and trend over time.
The multi-touch journey problem
Even when you can track that a user came from perplexity.ai, there's a second problem: the B2B purchase journey rarely starts and ends on the same channel.
The most common scenario: the customer discovers your company via ChatGPT but doesn't convert immediately. Days later, they search your name on Google, click the organic result, and convert. The last-click attribution model credits Google. The first-click model credits direct (because ChatGPT didn't generate a trackable click). The AI doesn't appear as responsible for the conversion in any standard model.
That's the attribution problem no analytics system fully solves today. The available approaches are imperfect but useful:
Origin survey with new customers. Include in your onboarding process the question "how did you hear about us?" with AI (ChatGPT, Perplexity, etc.) among the options. Self-reported data has bias, but it's surprisingly informative for understanding the channel mix — especially for origins analytics doesn't capture.
AI traffic-segment analysis. Compare the AI-referral segment's quality metrics (time on site, pages per session, sign-up conversion rate) with other channels. Even without precise revenue attribution, the traffic-quality difference justifies investing in the channel.
Temporal correlation. If you ran an AI-visibility action in March (published a press release, landed a story in an authoritative outlet) and in April direct traffic and leads rose, the temporal correlation is evidence — weak, but real — of causality.
The argument for investing even without perfect tracking
The lack of perfect tracking isn't an argument not to invest — it's an argument to calibrate expectations and use alternative metrics.
The fastest-growing companies in digital marketing rarely have perfect tracking of every channel. SEO, today the most mature channel in digital marketing, took years to have acceptable attribution models — and even now, "organic traffic" imperfectly captures all the influence content has on the purchase decision.
The AI channel is at the same moment SEO was at in 2010: growing fast, hard to measure precisely, but clearly generating results for those who invest. The companies that accepted the ambiguity and invested in SEO in 2010 reaped competitive advantages that persist to this day.
How Crowly helps close the attribution gap
Crowly doesn't solve the revenue-tracking problem — no tool fully solves it today. What Crowly offers is the mid-funnel metric that connects content and presence effort to measurable visibility: a citation score per platform, week-over-week trends, a comparison to competitors.
That score is the most reliable available proxy for "how much AI investment is generating results" — even if it doesn't reach revenue directly.
Build AI visibility now, before the tracking is perfect. Companies that wait for perfect tracking arrive at the channel late. Free diagnostic →
Sources:
