Measure clicks from AI search results with Google Tag Manager

Tracking Google AI Overviews in GA4: setup outdated since May 2026

Ralph Grundmann · Last updated on 5 November 2026

Outdated: as of May 2026, checked in November 2026
This setup measures clicks where a text fragment (#:~:text=) is appended to the target URL. Since the update in early May 2026, this no longer works for AI Overviews. Google now delivers these links as inline links without a text fragment. Our own check on 5 November 2026 confirms this: the source links in an AI Overview go through a Google redirect to the target page, and none of the links we tested arrived with a text fragment.
Where the setup still applies: for “People Also Ask” and classic Featured Snippets, Google continues to link with a text fragment, so these clicks are still recorded. The instructions below use the event name text_fragment_click for them.

New in GA4: Since 13 May 2026, GA4 automatically reports traffic from AI assistants (ChatGPT, Gemini, Claude, among others) in a separate “AI Assistant” channel within the Default Channel Group, entirely without a specific setup (What’s new in Google Analytics, 13 May 2026). Important: This channel captures chatbot referrals, not Google AI Overviews, which continue to count as Organic Search.

New in Search Console: since 31 August 2026 Google has offered a dedicated generative AI report worldwide. It shows how often your pages were displayed in AI Overviews and AI Mode. It does not include clicks or queries. You can read what the report can do and where its gaps are in our article Search Console now shows AI visibility.

Since Google rolled out AI Overviews in search, the way users reach websites has been changing. AI-generated summaries with further links to external sources appear in the search results. This raises a very practical question for many SEO, content, and analytics teams: How can traffic from Google AI Overviews be made visible in GA4?

There is no direct standard solution for this in Google Analytics 4. Until May 2026, such visits could be tracked with a pragmatic setup in Google Tag Manager. This no longer works for AI Overviews. The following setup still records visits with a text fragment, for example from “People Also Ask” and Featured Snippets. The basis is a so-called text fragment in the URL (#:~:text=). This technique allows browsers to jump directly to a specific point on a page and highlight it. Text fragments are part of the URL fragment, are processed client-side and never reach the server.

This behaviour can be used for tracking: When a user lands on your website via a link with a text fragment, a GTM setup can recognise the passed excerpt and send it as an event to GA4.

Measurable visits

What can be measured with AI Overviews

It is important to be clear about what this measures: this setup does not measure whether an AI Overview has been displayed in Google search. It exclusively measures whether a page view on your website has arrived with a text fragment.

This is a crucial difference. Therefore, you will not receive a complete impression report for AI Overviews, but a very useful measurement for specific visits, which arrive with a highlighted text snippet on your page. Google describes AI Overviews as part of its AI-powered search features, which assist users with more complex queries and contain links to web content.

Measurable data at a glance

Which data the setup makes visible in GA4

With this approach, the following information can be evaluated in GA4:

whether a visit occurred with a text fragment
which target page was accessed
which excerpt in the fragment was passed (standard GA4 properties limit parameter values to 100 characters)
how often such visits occur

This is particularly helpful if you want to understand which content is referenced in AI contexts and which landing pages receive additional traffic as a result.

Limits of classical measurements

Why normal referrers fail with AI traffic

Clicks from AI Overviews come from Google Search like classic results and count as Organic Search in GA4, so the referrer cannot separate them from classic search results. Until May 2026, text fragments provided a practical approach here, as they are directly identifiable in the target URL. Since fragments are processed on the client side, they are not suitable for server-side evaluation, but are very well suited for a client-side GTM setup.

The technical structure

This is how tracking works with GTM and GA4

The setup consists of three components:

a custom JavaScript variable
a page view trigger
a GA4 event tag

The variable checks whether a text fragment of type #:~:text= is present. If so, the snippet is extracted and processed. Subsequently, a trigger fires on a page view basis, and a GA4 event sends the information to your property.

Extract text fragments

1. Custom JavaScript variable for the text fragment

First, create a new variable in Google Tag Manager.

To do this, navigate to Variables → New → Custom JavaScript in your container.

It is best to assign the following name to the variable:

Recommended variable name:
JS – Text Fragment

Next, insert this code:


function() {
  try {
    // Retrieve all performance entries of the current page view
    var entries = performance.getEntries();

    // Collect matches with text fragment
    var matches = [];

    for (var i = 0; i < entries.length; i++) {
      if (entries[i].name && entries[i].name.indexOf("#:~:text=") !== -1) {
        matches.push(entries[i]);
      }
    }

    // No text fragment found
    if (matches.length === 0) {
      return null;
    }

    // Extract text fragment from the URL
    var rawMatch = matches[0].name.match(/#:~:text=(.*)/);
    if (!rawMatch || !rawMatch[1]) {
      return null;
    }

    // Take the first text directive, drop prefix (ends with -) and suffix
    // (starts with -), decode start and end (separated by a comma) one by one
    var fragment = rawMatch[1].split("&")[0].split(",")
      .filter(function (part) {
        return part.slice(-1) !== "-" && part.charAt(0) !== "-";
      })
      .map(decodeURIComponent)
      .join(" … ");

    // Return to GTM
    return fragment || null;

  } catch (e) {
    // Do not send values on errors
    return null;
  }
}

Create Fragment Trigger

2. Trigger: page view with snippet detection

Create a trigger in GTM now.

Recommended Name:
PV – Text fragment detected

Trigger Type:
Page View or Window Loaded

Condition:
{{JS – Text Fragment}} is not equal to null

This way, the trigger will only fire if a text fragment has been detected.

Send data to GA4

3. Create GA4 Event Tag

Recommended tag name:
GA4 – Event – text_fragment_click

Event name:
text_fragment_click

Event parameters:

ai_snippet_text = {{JS – Text Fragment}}

landing_page = {{Page URL}}

Why these names make sense:

For custom GA4 events there are clear naming rules: Event names must start with a letter and may only contain letters, numbers, and underscores. Furthermore, consistent snake_case is significantly easier to maintain in practice.

Consent Check

4. Respect consent

The event may only be sent to GA4 if the user has consented to Analytics. In GTM, the tag should therefore be linked to your existing consent logic, for example via analytics_storage.

Linking the tag to consent is mandatory.

5. Create custom dimensions in GA4

To use the values passed later in explorations or reports, you should register the event parameters as custom dimensions in GA4.

Recommended:

ai_snippet_text
landing_page

After that, you can analyse which URLs were accessed most frequently through this event and which text snippets were passed along with it.

Final check

6. Test setup

Check the setup with the GTM preview and the DebugView in GA4.

Important test points:

Does the variable return a value?
Does the trigger fire only when the text fragment is present?
Does the event arrive in GA4?

Are ai_snippet_text and landing_page passed correctly?

Limits of Tracking

What the setup cannot achieve

This tracking is deliberately pragmatic and has clear boundaries.

It only measures page views with text fragments. It does not record impressions from AI Overviews. Furthermore, the setup depends on the presence of a corresponding fragment in the target URL that is recognisable on the client side. Not every visit from an AI interface must necessarily be passed in this way. Conversely, the event counts every visit with a text fragment, including those from “People Also Ask”, Featured Snippets or other websites.

Nevertheless, the approach is valuable in practice: it creates transparency in an area where classic referrer data alone often does not help.

Visibility & Visits Combined

Evaluate SISTRIX and GA4 data together

A useful extension of the setup is to compare it with SISTRIX data. The AIO analysis in SISTRIX shows for a domain which keywords Google displays an AI Overview for, which URL is cited and at which position it appears within the box, and whether the cited URL also ranks organically.

Additionally, metrics such as search volume, competition and CPC are available. On this basis, the evaluation in GA4 can be better contextualised: SISTRIX provides information about the fundamental presence of a domain in AI Overviews, while GA4 shows which target pages receive visits from Organic Search. By combining both data sources, it can be distinguished whether a URL primarily benefits from its AI Overview presence or already performs strongly organically regardless. This enhances the assessment of topic clusters, landing pages, and optimisation potentials in the context of AI Search.

Contact

Do you want to classify AI traffic in GA4 reliably?

Whether it’s GTM setup, event concept, consent mode or evaluation in GA4: we help you capture visits from AI assistants and Google Search reliably and put them into context. Speak to us if you would like to review, expand, or implement your tracking setup directly.