More and more B2B buyers start their research by asking an AI assistant. They ask ChatGPT for a shortlist of tools. They ask Claude to compare two vendors. They ask Perplexity what other companies use for a problem. Then some of them land on your website.
Here is the problem. When those buyers arrive, your analytics often has no idea where they came from. Some visits carry a small label that names the AI tool. Many carry nothing and get filed under Direct. And the biggest group never clicks at all. The influence is real, but the report cannot see it.
This guide shows you where AI visits hide in your reports, how to make the visible part show up properly, and how to measure the part that never produces a click.
Why AI Traffic Hides in Your Reports
When someone clicks a normal link on the web, their browser usually tells your site which page the click came from. Analytics tools use that information to group visits into channels like organic search, social, and referral.
AI assistants break this system in three ways.
First, some AI links carry a label and some do not. ChatGPT often adds a tag called utm_source=chatgpt.com to links people click in the browser. Those visits can be found and grouped. But a click from the ChatGPT mobile app usually arrives with no label at all.
Second, unlabelled visits get filed under Direct. When a visit arrives with no source information, analytics tools assume the person typed your address in by hand. So a growing slice of your direct traffic may actually be AI traffic wearing a disguise.
Third, many AI answers end without any click. A buyer reads what the assistant says about you, closes the tab, and comes back days later by searching your name on Google. The credit lands on branded search or Direct. The AI conversation that started the journey leaves no trace.
These patterns change as the AI tools update how they open links. Treat this table as a starting point and check what actually appears in your own reports.
A buyer can ask an AI tool to compare five vendors, read the answer, and choose you without clicking a single link. Your analytics records nothing. Often the only trace this moment leaves is what the buyer tells you later.
Step 1: Group AI Visits in Google Analytics
Start with the visits you can see. In Google Analytics you can create a custom channel group so that AI visits stop hiding inside Referral and Direct.
In GA4, open Admin, then Data display, then Channel groups, and create a custom group. Add a new channel called AI and set its condition to match visits whose source contains any of the AI domains: chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com, and gemini.google.com. Put the AI channel above Referral in the list, because GA4 uses the first rule that matches.
For a quick look without any setup, open your traffic report, change the dimension to Session source, and search for chatgpt. Anything you find there is the visible tip of your AI traffic.
Keep one thing in mind. This step only counts visits that arrived with a label. The app clicks that show up as Direct, and the answers that were read but never clicked, are still invisible. That is what the next step is for.
Step 2: Ask Your Buyers Directly
The most reliable way to see AI influence is also the simplest. Add one open text question to your sign-up or demo form: "How did you hear about us?"
People answer this question honestly and specifically. Buyers who found you through an AI tool tend to say so in plain words, such as "I asked ChatGPT for tools that do X and it suggested you". No tracking script can capture that moment. The buyer can, and will, if you ask.
We wrote a full guide on how to set this question up well, when to ask it, and what to avoid. You can read it here: the complete guide to How Did You Hear About Us surveys.
As answers come in, group the AI mentions into their own category, next to channels like podcasts, events, and recommendations. You can do this by hand in a spreadsheet while volumes are small, or use a tool that reads and classifies the answers for you once they grow.
Step 3: Put the Two Views Side by Side
Now compare what your analytics sees with what your buyers say.
Here is a made-up example to show the pattern. Suppose your new AI channel shows two percent of visits coming from AI tools. At the same time, one in five new customers writes that they first found you through ChatGPT. Both numbers are invented, but the shape of the gap is common: click data understates AI influence, because so much of that influence never produces a click.
The gap between the two views is not an error to fix. It is the measurement. It tells you how much of your AI influence is invisible to analytics, which assistants your buyers actually use, and whether that influence is growing quarter by quarter. This is the same approach we recommend for the wider set of private, untracked influences known as the B2B dark funnel.
How to Show Up Well in AI Answers
Once you can measure AI influence, the natural next question is how to earn more of it. The full answer deserves its own guide, but the starting point is simple: AI assistants read public web pages, and they favour pages that state facts clearly.
Keep your pricing, features, and comparisons on public pages rather than behind forms. Answer the questions buyers actually ask, in plain words, close to the top of the page. Keep your facts consistent across your site, because assistants notice contradictions. Pages written simply for people turn out to be the same pages assistants quote.
Start Counting Before It Gets Big
AI assistants are becoming a normal first step in B2B buying research. The teams who set up an AI channel and an open text survey question today will have a year of trend data when their board asks about it.
Neither setup takes more than an afternoon. One makes the visible part of AI traffic countable. The other reveals the part no analytics tool will ever see.
Measure it now, while it is small enough to watch grow.
Frequently Asked Questions
What does utm_source=chatgpt.com mean in my analytics?
It means someone asked ChatGPT a question, ChatGPT mentioned your website in its answer, and the person clicked through. ChatGPT adds that label to many of the links it shows, which is why the visit arrives with a source you can group and count. Treat it as a clear signal that your site is being mentioned in AI answers.
Why does AI traffic show up as direct traffic?
When a person clicks a link in a mobile app, or copies a link and pastes it into their browser, the visit arrives with no information about where it came from. Analytics tools file these visits under Direct. Many AI assistant visits travel exactly this way, so a rise in direct traffic can hide a rise in AI influence.
How do I know if ChatGPT is recommending my product?
Two simple checks work well together. First, ask new customers how they heard about you with an open text question, and watch for answers that name an AI tool. Second, ask the assistants the questions your buyers would ask, such as "what are the best tools for X", and read what they say about you and your competitors. Repeat this monthly, because AI answers change often.
Is AI traffic worth treating as its own channel?
Yes. For most B2B companies it is still a small share of visits, but it is growing, and the people who arrive this way have usually done real research before they click. Grouping it now gives you a baseline, so in six months you can see the trend instead of guessing.
Can I track AI traffic without buying new software?
You can get a long way with what you have. A custom channel group in Google Analytics makes the visible part of AI traffic show up in your reports, and one open text question on your forms reveals the invisible part. Software helps once the volume of survey answers becomes too large to read and sort by hand.
James Kevan is the co-founder of First Signals, which reads open text answers to "How did you hear about us?" and turns them into attribution evidence, including mentions of AI assistants that analytics tools cannot see.
Related guides: The B2B Dark Funnel · HDYHAU Surveys for B2B · Why Your Attribution Dashboard Is Lying To You
© 2026 James Kevan / firstsignals.ai. Share freely with attribution.