IA Traffic: Measure in GA4 and Convert to Leads with UTMs

You may have experienced this scene: your site appears in a response from ChatGPT, Perplexity, or Copilot, you click, you rub your hands... then you open GA4 and find yourself facing a great artistic blur. Part of the traffic falls under Referral, another under Direct, and regarding leads, you don’t really know what comes from where. It’s not that AI traffic “doesn't work.” It's mainly that it is poorly attributed and often poorly welcomed. And that’s a shame, because these visitors don’t arrive by chance. They have already asked a question, already eliminated options, and generally click to validate information, compare a solution, find an example, or take action.

The goal here is simple: to help you properly measure this traffic with GA4, to sanitize attribution when you control the links through UTMs, and to convert these visitors into leads with a strategy of “source” pages that make people want to click, stay, and make a request.

Why AI traffic is different and why you shouldn’t treat it like classic SEO

When we talk about AI traffic, we refer to visits that come to your site after a user has clicked on a link provided by an assistant. Depending on the platform, this click may come from a web interface, a mobile application, or an integrated browser. It can also go through intermediate redirects. The result: in GA4, attribution isn’t always clear and you may feel like “it’s not coming through.”

What makes this traffic interesting isn’t necessarily its volume, especially at first. What makes it valuable is intention. The person doesn’t stumble upon your page casually. They have formulated a need, received a partial answer, and they often click because they want to confirm, verify, compare, or act. If you only analyze the number of sessions, you miss what matters: the quality of engagement and your pages' ability to convert.

The most common mistake is to let GA4 categorize this traffic wherever it can, then to give up due to lack of visibility. Another mistake is measuring only the final form submission, without following the micro-signals that indicate a visitor is progressing. Finally, many sites welcome these visitors on too vague pages, which don’t clearly propose the next step. With such intent-driven traffic, it’s a missed opportunity.

Preparing GA4 to track AI traffic without overcomplicating it

Before trying to isolate “ChatGPT” in your reports, you need to ensure that your conversions are clean. Otherwise, you will see noise instead of measuring real performance. It is essential to define what you consider to be a lead, and to distinguish the main conversion from what is an intermediate signal.

In many cases, a request for a quote or a meeting booking are the final conversions. But clicks on a key button, downloading a template, or newsletter sign-ups are signals that indicate real intent, especially when they come from a page that answers a specific question.

In GA4, the idea is not to stack dozens of events. It’s better to have a clear and reliable set of events that you can relate to business goals. Once these events are in place and tested, you can mark those that count as conversions in GA4. This foundation will then allow you to compare sources with each other, including AI.

Identifying AI traffic in GA4 and stop guessing

To begin, you can go to the acquisition reports in GA4 and observe the session sources and mediums. What becomes really useful is cross-referencing this information with the landing page, because AI traffic is very dependent on the page the person lands on. You may have low but highly profitable AI traffic on one page, and bigger but less qualified traffic on another.

The most practical step is to create a dedicated exploration in GA4. The idea is to display a table that shows the session sources and mediums, the corresponding entry pages, and the associated conversions. You can then apply a filter based on what you observe in your referrers. There’s no need to aim for encyclopedic perfection. What matters is identifying the reality of your site: where these visits come from and which pages actually convert.

To take it further and avoid doing this work manually every time, creating a custom channel “AI Assistants” is a real time-saver. A dedicated channel allows you to more easily compare AI traffic to organic, direct, or paid traffic. You get a readable dashboard, and you can finally answer the useful question: does this channel bring leads, and under what conditions?

UTMs: Sanitize attribution when you control the link

The UTMs will not solve the entire problem because you do not control the links that AIs display spontaneously. However, as soon as you master a link, they become very valuable. Whenever you share a template, a resource, a PDF, a newsletter, a signature, or a campaign, you can tag the links and obtain more stable attribution. This is particularly useful when click paths go through redirects or environments that obscure the original source.

To avoid inconsistent analyses, you need to choose a simple UTM convention and stick to it. A clear structure with a source, a medium, a campaign, and possibly a content tag is more than sufficient. The goal is not to add details for the sake of it, but to group and compare in GA4 without creating a hundred different variations for the same reality.

The most important point, the one that makes the difference between “I have UTMs” and “I can manage,” is the ability to capture this information in your forms. If you don’t capture the source and campaign when a person contacts you, you lose the connection between the analysis and the lead. With hidden fields in your forms, you can store UTM parameters as well as the landing page, and then cross-reference with your requests, your CRM, or even a simple tracking table.

“Source” pages: The most profitable strategy to capture and convert AI traffic

Measuring is good. Converting is better. And this is where “source” pages become a very effective tool. A source page is designed to be understood, referenced, and clicked. It clearly answers a specific question, structures the response with explicit headings, and provides what AIs love to cite: a clear definition, a method, steps, examples, and a FAQ that aligns with the wording users type in their prompts.

Some pages are particularly well-suited for this logic. Complete guides, checklists, templates, comparisons, and glossaries have a strong click-power because they promise immediate utility. When a reader from an AI clicks, it’s often to save time and secure their decision. If your page gives them an actionable answer, they will stay. If you additionally offer a clear next step, they will proceed.

The structure of a converting source page is based on a simple idea: you need to make the page scannable and reassuring. An introduction that announces the outcome, a quick summary of key points, a table of contents, short sections with concrete examples, and a FAQ oriented towards “questions” rather than “chitchat.” Then, you add visible and contextualized calls to action. A button lost at the bottom titled “Contact” is rarely sufficient. However, proposals like “Download the template,” “View a filled example,” or “Quick audit” create a natural continuity with the AI approach: the person wants to move from “I understand” to “I do.”

Transforming AI visits into leads with a simple journey and minimal friction

AI traffic doesn’t require a complex funnel. It requires clarity. The person arrives on a page that answers their question, sees a useful resource, leaves their email or requests an audit, and then proceeds to a service page or a conversation. In many cases, the conversion happens in two steps: a first light engagement, then a stronger request. If you only offer direct contact, you risk losing part of the visitors who are still in a validation phase.

The message must also be tailored to the visitor's mindset. They arrive with a partial answer, not a blank page. They want to confirm that you know what you’re talking about and that they can trust you. They also want to feel that you have a method, not just a promise. The goal is not to write “more,” but to write “better,” with elements that reduce uncertainty: a framework, an example, a checklist, a way to measure.

Managing quality: Don't settle for just how much

The classic risk with a new channel is to stop at volume. With AI traffic, you must look at engagement and conversion per page. A page may receive few sessions and yet generate very qualified leads. This is even common because AI clicks often align with specific queries. Therefore, it is useful to compare the engagement rate, conversions, and performance of each entry page.

If you have the ability to link GA4 to your CRM or tracking system, even in a simple way, you will take a leap forward. Knowing whether leads from this channel convert more often into customers allows you to prioritize the creation of source pages, improve CTAs, or produce more relevant lead magnets. You can also set up a very simple scoring system based on intent, profile, and action taken, to identify which pages attract the best contacts.

A realistic action plan for one month, without becoming obsessive

In the first month, the goal is to clean up the tracking, observe what GA4 truly reports, and create an initial source page designed to capture and convert. Then, you adjust based on the data, improving what underperforms and strengthening the internal linking to your service pages. You regularly compare the performance of the AI channel with that of other channels, not to declare a winner, but to understand which formats and pages create leads.

AI traffic is an opportunity, but only if you manage data and experience

AI traffic is a hybrid channel. It can be blurry to attribute, but it is often very intent-driven. To leverage it, you need to ensure that your conversions are measured correctly, that you can isolate this traffic in GA4 through a dedicated exploration and channel, that you use UTMs when you control the distribution, and that you build “source” pages that entice clicks and offer an obvious next step.

If you want to turn this channel into leads, ask yourself one last simple question: when someone arrives from an AI to my site, am I immediately offering something useful and concrete to move forward? If the answer is “yes,” you are on the right track. If the answer is “meh,” that’s precisely where you have the biggest room for improvement.

What you can implement right now, without redoing all your tracking

If you want to move quickly, think “minimum viable tracking.” The idea isn’t to set up a complicated system, but to make your data clean enough to make decisions. Start by checking that your main conversions are indeed coming through in GA4, and that you have at least one or two reliable intermediate signals. If you only measure the final form submission, you will always feel like “it’s not converting,” while in reality, you’re missing the journey.

Next, get used to looking at your AI traffic from the “landing page” perspective. The common reflex is to search for a perfect source, a clean label, a clear referrer. Except that this channel isn’t always clean. However, your pages are. If you identify that a specific page attracts visitors from assistants and triggers useful actions, you already have actionable information: you can improve this page, link it to your services, and create sister pages on related questions.

Finally, whenever you control a link, think UTM. Not to “look nice,” but to link the effort (distribution, resource, newsletter, template) to a result. And if you really want to manage, capture these parameters in your forms. It’s often this little detail that transforms GA4 into a decision-making tool instead of just a statistics screen.

When you don’t have control over AI links, the real weapon is content

There’s a common frustration with AI traffic: you can’t force an AI to display your link. And even if it does, you don’t always have control over the exact context. That’s true. But you have enormous power: you can create pages that become naturally “quotable.”

A quotable page is not a “long” page. It’s a page that is clear, structured, and precise. It starts by quickly answering, without beating around the bush. It defines terms. It explains a method. It shows an example. It anticipates objections. And it ends with a logical next step. AIs love to reference this type of content because it’s easy to summarize without misrepresenting it. Humans love it for exactly the same reason.

In other words: if you build good source pages, you reduce your dependence on the “perfect referrer.” You transform traffic that is sometimes difficult to attribute into traffic that is much easier to convert. And when it converts, attribution becomes less anxious, because you finally manage what really matters: the result.

Make your “source” pages more compelling than the AI response

The challenge is simple: when a visitor comes from an assistant, they often arrive with a partial answer. They’re not coming to reread what they already know. They come to get something better: more concrete, more reliable, more actionable. If your page just rephrases the generic answer they just read, you lose them.

What works is to add what the AI doesn’t always provide well at that moment: clear context, a realistic course of action, truly usable examples, and benchmarks for making decisions. In practice, a “source” page that converts often resembles an effective conversation: you respond, you clarify, you show, and you propose a simple continuation.

And this continuation must be obvious. If your only call-to-action is “Contact,” many visitors won’t be ready. However, if you offer a useful resource, a quick audit, a filled example, or a mini-method to download, you capture a lighter engagement. It’s often this little “first yes” that opens the door to the lead.

The questions you will be asked (or that you will ask yourself) as soon as you look at AI traffic

Why does GA4 often classify this traffic as Direct or “weird” Referral?

Because AI click paths sometimes go through environments that don’t always transmit a clean referrer, or that use redirects. Depending on the interface (web, mobile, integrated browser), the source may be partially lost. Hence the importance of working by entry pages and using UTMs whenever you control the distribution.

Do I absolutely need to isolate “ChatGPT” and “Perplexity” in GA4?

It’s useful, but not mandatory at first. If you can already isolate a “AI Assistants” group in a dedicated channel and link this traffic to pages and conversions, you can already manage. Fine granularity is a bonus, not a condition for achieving results.

What volume of AI traffic can we expect?

Often, it's not a tidal wave at the beginning, and that's normal. The signal is more “qualitative” than “massive”. What matters is the ability to convert strong intent. A page that receives few sessions but generates qualified requests is often worth more than a thousand cold visits.

What if I see almost nothing in GA4?

Start by checking that your conversions are properly configured, that you're looking at the right reports, and that you're analyzing by landing pages. Then, create one or two “source” pages really designed to be cited and clicked. Very often, AI traffic becomes visible when you give assistants a good reason to send you visitors.

AI traffic is not a mystery; it's a channel to tame

Yes, the attribution of AI traffic can be blurry. Yes, GA4 can give you the impression that everything is sorted randomly. But this channel has a rare advantage: intent is often already mature. Visitors do not arrive by accident; they come to validate information, compare, or take action.

If you want to leverage it, keep a simple logic: clean tracking to measure what matters, UTMs whenever you control a link, and above all, “source” pages designed to be understood, cited, and clicked. From there, you no longer endure the channel: you manage it.

And if you want to go further without spending your evenings on it, you can also get support in creating source pages, optimizing editorial content and SEO, or structuring your content plan. This is exactly the kind of project where a writing team can save a lot of time while maintaining quality and conversion standards.