Google AI Overview Link Tracking Parameters: What To Do

4 October 2026 · The Cresia team

If Google's test becomes a standard feature, clicks from AI Overviews will be identifiable in your analytics instead of hiding inside general Google organic traffic. That's a real improvement for measurement. It won't tell you how often you're cited, quoted or recommended without a click, so treat it as better plumbing, not a new strategy.

Key takeaways

Search Engine Roundtable reports that Google appears to be testing tracking parameters on links inside AI Overviews, which could let you separate those clicks from ordinary web-result clicks.
It's a test. Get your analytics ready to catch the parameter, but don't reorganise your AEO programme around it yet.
A tagged click measures what happens after a visit. It says nothing about the citations and mentions that never produce one.
Set up landing-page segments, log retention and a baseline now, so a before-and-after comparison exists if the change ships.
Keep citation tracking and click tracking as separate measures, because they answer different questions.
Don't strip, rewrite or redirect the parameters, and don't build reports on one until you've seen it in your own logs.

What Google is reportedly testing

Search Engine Roundtable reports that Google seems to be adding tracking parameters to the link URLs inside AI Overviews. If the test rolls out fully, a publisher could tell whether a visit came from an AI Overview or from a regular result on the same results page.

That's the whole of what's been reported. It's early. A test can be limited to certain users, queries or countries, and it can change format or disappear. Nobody outside Google can say today whether it will cover AI Mode too, or how stable the parameter will be.

So the sensible first move is dull. Check your own landing-page URLs and server logs for anything new appended to Google-referred visits, and read Google's Search Central documentation for what's officially supported. Don't build anything on a parameter name you've only read about in a news post.

There's a second reason to stay measured. Search Engine Roundtable also reported a Google Ads sitelinks format appearing in AI Mode. Taken together, the two stories say the same thing: Google is treating AI surfaces as places with their own link formats, their own ad units and, possibly, their own attribution. The surfaces are being productised. Your reporting should expect to follow.

Why a separate AI Overview click matters for AEO and GEO

Until now, most teams have had a vague sense that AI answers affect their traffic and very little evidence of how. A visit that arrives from an AI Overview and a visit from a standard blue link can look the same in a session report. Both say google / organic.

That has real consequences for decisions. Picture an in-house SEO lead at a software company who sees organic sessions on a comparison page fall eight weeks running. Is the page losing rankings? Is an AI Overview answering the question above it? Is it still being cited but clicked less? With no way to split the traffic, the honest answer is a guess, and the guess usually follows whoever argues loudest in the meeting.

A tagged click doesn't settle all of that. It does give you one clean number per page: visits that came through an AI answer. Put that beside rankings and total organic sessions, and some of the guessing stops.

It also changes the conversation about value. If AI Overview visitors convert differently from standard organic visitors, you'll be able to see it. They might convert better because they arrive pre-informed. They might convert worse because the answer already satisfied most of the need. I wouldn't predict which. The point is that you'll be able to look instead of arguing from anecdote.

How AI answer engines choose and cite sources

A tracking parameter tells you where a click came from. It doesn't tell you why your page was linked, and that's the question AEO and GEO work is really about.

What's reasonably well established, without leaning on any single study: answer engines favour pages that state a clear answer to a specific question, in a passage that still makes sense when lifted out of the page. They favour pages from sources that look credible for the topic, and pages that are easy for their crawlers to fetch and parse. Retrieval systems tend to pull passages, not whole pages, so a page with one strong, self-contained paragraph can beat a longer page that never quite commits.

A few practical consequences follow:

Passages matter more than pages. Each section should answer one question completely, with the answer in the first sentence or two.
Wording matters. Search Engine Journal ran a piece on finding the exact phrases that get content cited in AI search. The underlying idea is sound: match how people phrase the question, and how the answer is usually phrased, instead of inventing your own vocabulary.
Structured data is table stakes, not a trick. Search Engine Journal's Ask An SEO column recently covered structured data mistakes that hurt AI visibility. Broken or contradictory markup is a cheap thing to fix and an expensive thing to leave.
Entity clarity helps. If your company, product and authors are described consistently across your site and elsewhere, engines have less to reconcile.

None of this is secret, and none of it depends on the tracking test. The test only affects how well you can see the results.

What to change in your analytics now

You can't act on a parameter you haven't seen, but you can make sure you'll notice it and use it. Here's a short sequence that works whether the test ships next month or never.

Take a baseline today. Export 90 days of Google organic landing-page sessions, conversions and Search Console clicks and impressions for your top 50 pages. If a split appears later, you'll want something to compare it against.
Check what your analytics does with unknown query strings. Some setups drop them, some keep them, and some store them in a way that's hard to report on. Find out which yours is before a new parameter arrives.
Keep raw server or CDN logs long enough to look back. Logs show the full URL a visitor requested. If a parameter appears, your logs will show it before your dashboards do.
Make sure canonical tags and redirects behave. A parameterised URL should resolve to the same page, canonicalise to the clean URL, and not trigger a redirect that strips the evidence. Test this with a made-up parameter on a staging page.
Write the parameter into your tracking specification once it's confirmed. Naming it, defining it and saying how it maps to a channel grouping keeps three teams from inventing three definitions. What is a tracking specification covers the idea, and OmniSpec is the place to look if analytics governance is the gap.

Steps one to four take an afternoon. Step five waits for evidence.

A measurement plan that survives the test changing

The risk with any new signal is building a dashboard on it and then watching it vanish. A better plan treats the AI Overview click as one input among several, each with its own limits.

Question	Signal	Where to find it	Main limitation
Are we being cited in AI answers?	Manual and scheduled prompt checks, citation tracking	Your AI visibility tooling, or a documented prompt set	Answers vary by user, location and day
Do those citations produce visits?	Tagged AI Overview clicks, if the parameter ships	Analytics, server logs	Only covers Google, only clicks, only while the test lives
Are we losing clicks overall?	Search Console clicks, impressions and CTR by page	Search Console	Doesn't isolate AI features cleanly in standard reports
Does AI-sourced traffic convert?	Conversion rate by landing page and source segment	Analytics	Small samples early on; don't over-read a week of data
Is our brand being mentioned without links?	Branded search trend, direct traffic, mention tracking	Search Console, analytics, monitoring tools	Indirect; correlation, not proof

The second row is the one that may change. The other four exist already and will stay useful whatever Google does.

If you're building the citation-tracking side for the first time, MediaPilot is Cresia's product for media and AI-search visibility, and it's the place to start reading.

What a tagged click still won't tell you

This part matters most, and it's the part likely to be forgotten when a new metric arrives.

A click is the smaller share of what an AI answer does for you. Someone reads an Overview, sees your brand named as one of three options, and searches for you directly an hour later. That shows up as branded search or direct traffic. Someone else reads the answer, never clicks, and mentions your product to a colleague. That shows up nowhere.

A tagged-click report will make AI visibility look like a traffic channel. It's closer to a mix of awareness and consideration, with a traffic tail. If you judge your AEO work only by tagged clicks, you'll undervalue the pages that are cited often and clicked rarely, and you'll overvalue pages that happen to be linked prominently.

Two more limits are worth stating plainly.

First, the test is about Google. ChatGPT, Perplexity, Copilot and others each handle outbound links differently, and a Google parameter says nothing about them. Check each vendor's own documentation and referral behaviour instead of assuming they'll follow.

Second, a parameter is something Google chooses to add. It can be changed or removed without notice. Treat it as a convenience that might go away.

What not to do

When a new signal appears, teams tend to do something rash within a fortnight. A few things are worth avoiding.

Don't strip the parameter to keep URLs tidy. If your tag manager or CDN rules remove unknown query strings before your analytics sees them, you'll be blind to the thing you wanted to measure.
Don't build forecasts on the first week of data. Early AI Overview volumes may be small, uneven and affected by who's in the test. A fraction of a percent of sessions can swing wildly.
Don't write pages for the parameter. There's nothing to optimise in a tracking tag. The page still has to answer the question well.
Don't treat tagged clicks as the only AEO metric. Report them alongside citations, branded demand and conversions, not instead of them.
Don't rewrite your attribution model on a test. Add a segment, watch it for a quarter, and then decide whether it deserves a place in your standard channel grouping.
Don't announce results to leadership that you can't reproduce. A number from a feature that changes shape next month will be hard to defend.

Who needs to do what

This isn't only an SEO task. The parameter touches analytics, engineering and content, and it's easy for each to assume another has it covered.

Task	Suggested owner	Done when
Export baseline for top landing pages	SEO or analytics lead	90-day file saved with date and definitions
Confirm query strings survive to analytics	Analytics engineer	A test URL with a made-up parameter appears in reports
Check canonical and redirect behaviour	Web engineer	Parameterised URLs resolve and canonicalise correctly
Retain logs long enough to look back	Platform or DevOps	Retention period documented and confirmed
Add the parameter to the tracking spec	Analytics governance owner	Name, definition and channel mapping agreed
Keep citation checks running	AEO or content lead	A fixed prompt set is run on a schedule

Smaller teams will double up. That's fine. What matters is that the confirming step, the one that checks whether a parameter really appears in your data, has a named person.

If you're deciding how this fits with broader reporting, the analytics teams page lays out how Cresia approaches governance and measurement.

What I'd expect, and what I wouldn't

Here's a position. I'd expect Google to keep adding ways to tell its AI surfaces apart in reporting, because publishers have asked for years and because ads in AI Mode need their own measurement anyway. I wouldn't expect those reports to be complete. Platforms rarely hand over everything.

I also wouldn't expect this to change what earns a citation. Clear passages, sound structure, credible sources and fetchable pages did the work before the parameter and will after it. The teams who benefit most won't be the ones with the cleverest tag handling. They'll be the ones who had a baseline, kept their pages crisp and noticed the change early.

Frequently asked questions

Does Google's AI Overview tracking parameter exist for everyone yet?

No. Search Engine Roundtable describes it as a test, and Google hasn't presented it as a finished feature in anything we've cited here. Check your own logs for new query strings on Google-referred visits, and see Google's documentation for anything officially supported.

Will this tell me how often my site is cited in AI Overviews?

No. A tracking parameter only appears when someone clicks. Citations without clicks, brand mentions inside the answer and later direct visits won't show up. You'll still need separate citation checks.

Should I change my content because of this?

Not because of the parameter. The things that make a passage easy to quote, such as a direct answer, clear wording, valid structured data and a credible source, are worth doing anyway. The tracking change affects what you can see, not what gets chosen.

Will other AI search tools do the same thing?

That's unknown. Each vendor handles outbound links and referrals in its own way, and some send referrer information while others don't. Check each platform's documentation and your own referral data instead of assuming Google's approach carries over.

Is this worth acting on before it's confirmed?

The preparation is. Taking a baseline, confirming query strings survive, and keeping logs costs little and helps whether or not the test ships. Building reports on a parameter you haven't yet seen is the part to hold off on. If you'd like to talk through setup, you can request a demo.

Sources

https://www.seroundtable.com/google-ai-overview-link-tracking-parameters-42219.html
https://www.seroundtable.com/google-ads-sitelinks-ai-mode-42206.html
https://www.searchenginejournal.com/what-are-common-structured-data-mistakes-that-hurt-ai-visibility-ask-an-seo/589924/
https://www.searchenginejournal.com/where-to-find-the-exact-phrases-that-get-your-content-cited-in-ai-search/591562/