AI-Referred Visitors Need Their Own Conversion Line in the Report

Sooner or later a finance person asks the question: "What did all this AI visibility work do for revenue?" If AI-referred visits are buried inside "organic" or "direct" in the analytics, the honest answer is "we do not know". We would rather have the line.

This post is about giving AI-referred visitors their own place in the report and being straight about what that line can and cannot tell you.

Why standard analytics hides AI traffic

Standard analytics tools classify traffic by referrer. Some AI platforms pass a clear referrer, some pass an unusual one, and some pass none. Click-throughs from answers inside apps can land as direct visits. Result: AI-driven visits get scattered across channels, usually into direct or referral.

So the real number is not zero, and it is not what the default report shows.

What we set up before arguing about attribution

A tracking script that identifies AI sources. Promptwatch's visitor analytics counts AI-referred traffic and conversions from ChatGPT, Perplexity, Gemini, Claude, Copilot, DeepSeek, Kimi, Mistral, and others. It installs via a lightweight script, a Google Tag Manager template, or the WordPress plugin. We pick GTM for most clients, because it keeps the marketing team in control of deployment.

Conversion events. AI traffic only matters if it does something. We agree on the client's conversions: a demo request, a purchase, a signup. Then we make sure they are captured consistently.

A baseline. Start tracking before any program work if you can. Otherwise you cannot say what changed.

A cross-check in the client's own analytics. We compare the AI segment against what GA4 shows, understanding they will not match exactly. Different tools classify differently. The goal is to be in the same neighbourhood and to know why they differ.

How we present the line

Our monthly report has an AI traffic section with:

Visits from AI sources, by platform.
Conversions from those visits.
The pages that received them.
A comparison against other channels, only where the volumes make comparison fair.

We do not present AI traffic as a share of all traffic unless we explain what is in the denominator. We never extrapolate a single month into a forecast.

When the numbers are small

For many clients AI referrals are a small fraction of traffic today. We say so. A tiny line is not an argument against the work, and it is not an argument for it either. Several points help put it in context:

AI answers may influence decisions without sending clicks. A buyer reads the answer, remembers the brand, and later types the name into a search box. That shows up as branded search or direct. We note this possibility but do not claim a number for it.
Early data is thin. A handful of conversions swings percentages wildly.
Quality may differ from volume. Some published case studies, like Crisp's account that AI-referred traffic converted at twice the rate of traditional channels, point to higher intent, though that is that customer's result and not a promise for another site.

We cite such studies with their source and say they are one company's experience.

Linking traffic to pages and citations

Traffic by itself tells you little. We connect it back to pages and to the prompts that cite them. Promptwatch's 360 Insights joins crawler logs, citation clicks, and AI-referred visits per page. That lets us say things like: "These four pages are crawled, cited, and drive visits. These six are crawled and cited but drive none. These five are never crawled."

The zero-visit pages are interesting. They might need a stronger reason to click, such as a tool or a data table. Or the engine may be satisfied by quoting them in the answer.

Search Insights adds Google Search Console and Bing Webmaster data to the same view, including Copilot traffic from Bing, so the client can see AI and classic search together.

What we avoid claiming

That AI visibility directly caused revenue, unless the data supports a clear path.
That every mention converts.
That our numbers are complete. They are a floor, because some AI-driven influence does not leave a trace.
Percentages from tiny samples.

Handling the awkward meeting

If the client expected more, we show the full chain: visibility, citations, crawler access, visits, conversions. Each stage tells us where the leak is. Perhaps the brand is mentioned often but the pages are not cited. Perhaps the pages are cited but not visited. Perhaps visits arrive and do not convert because of landing page issues. Each has a different fix, and the report shows which one applies.

We make sure tracking respects the client's consent setup and local rules. The script should be configured the same way as other analytics tools on the site. If consent is not given, we do not collect around it.

Our recommendation

Give AI-referred visitors their own line from the first month, connect them to pages and conversions, and be candid about size and limits. We do this with Visitor analytics and 360 Insights in Promptwatch, which keeps the traffic, the crawls, and the citations in one view. You can read more at promptwatch.com.