How to Measure and Optimise AI-Driven Traffic in GA4

With the increased usage of language models (LLMs) and Google’s AI Overviews (AIOs), users are finding answers without ever visiting a website. For marketers and data analysts, this creates a worrying trend in reports: declining organic search traffic.

But declining volume doesn’t necessarily mean declining value.

This post is written for content marketing teams who need to look beyond vanity metrics and understand the real ROI of the new AI-qualified visitor. These are users who arrive after seeing an AI-generated summary and are often closer to taking action.

You’ll learn how to reclaim your data and understand whether AI tools are actually sending you high-value humans.

Google Analytics 4 (GA4) lumps traditional Google clicks and AI-driven visits into the same bucket. This hides meaningful behavioural differences and makes AI traffic hard to evaluate.

Create a Custom Channel Group to isolate AI referrals.

Include sources such as:

• Gemini
• ChatGPT
• Perplexity
• Claude

Reorder the list so AI Referral appears above the standard Referral group. This ensures GA4 checks for AI sources before defaulting traffic into a generic category.

Reference:
https://support.google.com/analytics/answer/13051316?hl=en#Example_AI_Chatbots&zippy=%2Cin-this-article

Why this matters:
If you don’t segment AI traffic, its behaviour and value get averaged out. What looks like “low impact” traffic may actually be high-intent users hiding in the wrong bucket.

Once you’ve isolated AI traffic in Google Analytics, the focus shifts to a single question:

Is your page satisfying the specific intent created by the AI?

Unlike a traditional searcher who clicks to discover an answer, an AI-referred user often arrives after seeing a summary and wants to act on the details. They aren’t looking for a “What is…” introduction. They’re looking for tools, examples, data, or next steps.

Look at how AI-referred users behave on your pages. In GA4, pay attention to these two patterns:

• Good match
  A high Engagement Rate combined with Key Event completions (for example, downloads) suggests your page delivered what the AI result promised.
• Something’s off
  Low Average Engagement Time paired with a high number of sessions usually means a disconnect. Users arrived with an expectation but couldn’t quickly find what they needed.

This step helps you identify which pages are aligned with AI-driven intent, and which ones create friction.

The real goal isn’t just showing up in an AI Overview. It’s capturing intent.

Because AI tools summarise your content upfront, users who click through are usually more motivated. They’re not browsing; they’re looking for the exact tool, template, or example you mentioned.

Start by comparing Session Key Event Rate across channels.

AI vs organic traffic
In GA4, isolate Session Key Event Rate by channel. You’ll often find that AI-driven traffic is smaller in volume but converts better than traditional organic search.

This happens because the AI acts as an early filter, sending users who have already seen your value proposition and are ready to act.

1. Navigate to Standard Reports → Traffic Acquisition
2. Use the channel grouping that includes AI Referral
3. Find Session Key Event Rate on the far right

If AI traffic is converting but not sticking around, the issue isn’t intent. It’s the page experience. That’s where page optimisation turns visibility into real results.

AI tools often send readers straight to a specific sentence, not the top of the page. If that section doesn’t clearly show what to do next, they’ll leave. This means you need to optimise for the entry point readers actually arrive on, not the page you originally designed.

Treat every high-value paragraph as its own mini landing page. Test small engagement elements to see what makes readers stay, click, or save instead of bouncing after reading the cited section.

Examples:

• Sticky banners
  Keep a clear “Next step” or “Download” button visible as people scroll.
• In-line prompts
  Add simple CTAs like “Save this,” “Get the checklist,” or “Read next” inside your strongest sections.
• Original visuals
  Use charts, calculators, or tables that add value and can’t be copied into an AI summary.

The goal

Don’t just answer the question that brought them there. Give them a clear reason and an easy path, to keep engaging.

Vicky Wong
Solutions Architect, FiveStones

Vicky works with marketers and analytics teams to design measurement frameworks, optimise GA4 implementations, and translate complex data signals into practical growth decisions. To learn about GA4, connect with Vicky at [email protected].