How to measure clicks from Google AI Overviews

AI Overviews sit at the top of a growing share of Google results, and publishers want one number: how many readers do they actually send? The honest answer is that nobody can count them exactly, but you can get a useful lower bound and a clear trend.

Why AI Overview clicks look like normal search

When someone clicks a source link inside an AI Overview, the browser sends the same google.com referrer as a click on a blue link. There is no separate referrer and no UTM tag, so GA4 and most other tools file the visit under Google organic search. AI Mode clicks are bundled the same way.

What Search Console shows

• The Performance report counts clicks from AI Overviews and AI Mode inside the normal "Web" search type, mixed with every other result. You cannot filter them out.
• The generative AI report, added in June 2026, shows how often your pages appear in AI Overviews, AI Mode and Discover's AI features, broken down by page, country, device and date. It shows impressions, not clicks or queries.

So Search Console tells you where you are cited. It does not tell you how many readers came from those citations.

The text-fragment signal

When Google links to a specific passage, it often adds a text fragment to the URL:

https://example.com/recipe#:~:text=bake%20for%2025%20minutes

The browser scrolls to that sentence and highlights it. Three limits matter:

1. It is not only AI Overviews. Featured snippets and "People also ask" use the same mechanism, so a fragment means "a passage-level Google feature", mostly AI Overviews today.
2. It is not on every click. Many AI Overview links have no fragment, so any count is a lower bound.
3. Servers never see it. The part after # stays in the browser. Server logs and many tag setups miss it unless a script reads it.

How to track it

In GA4 you need Google Tag Manager: a JavaScript variable that reads the navigation URL, a trigger when it contains :~:text=, and an event or custom dimension to store it. Note that Chrome removes the fragment from location.hash, so read it from performance.getEntriesByType('navigation').

In PubPulse nothing to set up. The tracker checks the navigation entry on the landing page. If the visit came from Google and the URL carried a text fragment, the visit is filed under the source Google AI Overview and counted in the AI traffic report next to ChatGPT, Perplexity, Claude, Gemini and Copilot. You can see it working in the live demo.

Reading the numbers

• Watch the trend, not the absolute. Because only part of the clicks carry a fragment, compare this month with last month on the same pages.
• Put it next to Search Console's AI impressions. Pages with many AI impressions and few fragment clicks are cited but not clicked: the answer already satisfies the reader.
• Look at engaged time. Readers who land on the exact passage often read less of the page. Check whether they go on to a second page; if not, link the next useful article right after the passage they were sent to.
• Compare with AI assistants. ChatGPT and Perplexity send a clean referrer, so they are measured fully. See how to track AI traffic.

What to do with it

Write passages that answer one question completely in two or three sentences, with the specific number or step the reader wants. Those are the passages Google quotes and links to. Then make the next step obvious, because a reader who arrives in the middle of your page has not seen your introduction.