Gemini UTM parameters: what they change for AI attribution

1 October 2026 · The Cresia team

Gemini traffic can now be told apart in your analytics, because Google has added UTM parameters to the links it sends people through, according to Search Engine Journal. That solves a reporting problem, not a visibility problem. You'll get cleaner session data, but you still won't see how often Gemini mentioned your brand without sending anyone to your site.

Key takeaways

Gemini's UTM parameters let you separate its clicks from generic referral and direct traffic. It's a reporting fix, not a visibility metric.
UTM tags only count visits that happened. They say nothing about mentions or citations that produced no click.
Answer engines choose sources at passage level, so better attribution won't change how often you're cited.
Check your own plumbing first. Redirects that drop query strings, internal UTMs and cache rules can wipe out the new data.
Report AI traffic, AI citations and brand outcomes as three separate lines, each with a named owner.

What Google changed with Gemini UTM parameters

Search Engine Journal reports that Gemini now appends UTM parameters to outbound links, so site owners can attribute visits to Google's AI platform more clearly. The piece was written by the outlet's Martin Ibuster. The practical effect is that a visit from Gemini carries a label you can read in your analytics tool, instead of arriving as an anonymous referral or as direct traffic.

That matters because AI-originated visits have been awkward to pin down. Some arrive with a referrer you can filter on. Some arrive with the referrer stripped and land in the direct bucket, where they sit beside people who typed your URL from memory. A tag on the link survives where a referrer header may not.

The exact parameter names and values, and which Gemini surfaces use them, belong in Google's own documentation. Read that before you write a single rule. Treat the whole thing as early: vendors adjust parameters, and a rule built on today's values can quietly stop matching later.

Why clean attribution matters for AEO and GEO

Take a hypothetical B2B software company. Its monthly report shows direct traffic climbing 14 percent while branded search is flat. Nobody can explain it. Part of the answer might be AI assistants sending people with stripped referrers. Without a tag, the team can't confirm that, so the growth gets credited to brand awareness and the budget conversation moves on.

With a reliable Gemini label, that story changes. You can ask whether Gemini visitors browse differently from organic visitors, whether they land on comparison pages or documentation, and whether they convert. Those are the questions that decide how much effort answer-engine optimisation deserves.

It also helps with an internal argument. Leadership often asks whether AI search is real traffic or a rounding error. A named, trended line settles that more quickly than any industry anecdote. Be ready for the answer to be small. For many sites AI referrals are modest next to organic search, and that's a legitimate finding, not a failure of the programme.

The larger value is still indirect. A person who asks Gemini for options, sees your brand named, and searches for you later never appears in the Gemini line. The UTM tag measures the click. The influence mostly happens elsewhere.

How AI answer engines choose and cite sources

Answer engines don't rank pages the way a results page does. Broadly, the system takes the question, often breaks it into several narrower lookups, retrieves candidate pages or passages, and then composes an answer from the pieces that fit. Sources that get cited are the ones whose passages answer a sub-question cleanly.

That has a few consequences for how you write and structure pages:

Passages beat pages. A section that fully answers one question can be lifted on its own. A page that circles its point for six paragraphs gives the engine nothing to quote.
Entities need to be unambiguous. If your product name, category and claims are stated plainly and consistently across your site and your profiles elsewhere, the engine can connect them.
Access comes first. If crawlers can't fetch the page, or the content sits behind scripts that never render, nothing else matters.
Structured data helps machines parse, not choose. Search Engine Journal's Ask An SEO column on common structured data mistakes that hurt AI visibility is a useful checklist for getting the markup right. Correct markup removes ambiguity. It doesn't guarantee a citation.
Phrasing matters. Search Engine Journal has also published a piece on finding the exact phrases that get content cited in AI search. The principle is sound: write the way your buyers ask the question.

None of this is touched by UTM tagging. Google adding parameters to outbound links changes what you can see after the click. It doesn't change retrieval, and it doesn't change who gets cited. Anyone selling UTM adoption as an optimisation tactic has the cause and the measurement backwards.

There's another point worth holding onto. Answer engines vary by query and by day. Search Engine Roundtable has covered AI Overviews showing up again for many large brand names, and for something as narrow as team rosters. Presence in these surfaces moves around. That's one more reason to measure over weeks, not days.

What to change in your analytics this week

Most of the work here is housekeeping, and it's the sort of housekeeping that fails silently. Go in this order.

Read Google's documentation for the exact parameters and the surfaces they apply to. Write the values down in one place.
Create a channel rule in your analytics tool that matches those values, so Gemini visits stop falling into default referral or unassigned groups. Test it against a real click from a Gemini answer, not just a hand-built URL.
Audit your redirects. Any redirect chain that drops the query string deletes the tag before analytics sees it. Check www to non-www, http to https, trailing-slash rules and any vanity URL redirects. This is the most common way attribution disappears.
Stop internal UTMs. If your own banners or navigation links add UTM parameters, they can overwrite the incoming source mid-session. Use event tracking for internal clicks instead.
Check your cache and tag setup. Make sure your tag manager reads the landing URL's query string, and that your CDN ignores UTM parameters in its cache key so each tagged URL doesn't become a separate cache entry.
Annotate the change. Mark the date the Gemini label began appearing in your reports. Without it, someone will read the step up in Gemini sessions as a growth story.

Write all of this into your tracking specification so it survives staff changes. If you haven't got one, what a tracking specification is covers the basics, and OmniSpec is Cresia's product for analytics governance.

What a UTM tag still can't tell you

The parameter answers one question: did this visit come from Gemini? Most of what AEO teams actually want to know sits outside that question.

Question you're asking	Does Gemini's UTM answer it?	Where to look instead
How many visits came from Gemini?	Yes, once your channel rule is in place	Your analytics tool
Did those visits convert?	Yes, if conversion tracking is intact	Analytics goals and CRM
How often was my brand named in answers?	No	Prompt sampling and AI-visibility tooling
Which of my pages were cited?	Partly, only where a click followed	Citation tracking, server logs
Did I lose share to a competitor in answers?	No	Repeated prompt tests over time
Did an answer influence a later branded search?	No	Branded search trends, survey questions

The gap in the middle rows is large. A brand can be cited constantly and see a thin Gemini referral line, because many answers satisfy the person without a click. The reverse can also happen. Treat the UTM line as the floor of your AI footprint, never the total.

A measurement plan that separates clicks from visibility

Keep three lines apart and give each one an owner. Mixing them produces the report nobody trusts: a single AI search number that blends visits, mentions and guesses.

Line	What you measure	Source	Owner	Cadence
AI traffic	Sessions, engagement and conversions from Gemini and other AI referrers	Analytics channel rules	Analytics lead	Weekly
AI citations	Share of a fixed prompt set where your brand or pages appear	Prompt sampling, visibility tooling	SEO lead	Fortnightly
Brand outcomes	Branded search, direct traffic, demo requests, self-reported source	Search Console, CRM, form fields	Growth lead	Monthly

Two habits make this plan hold up. First, fix the prompt set. Choose 30 to 50 questions your buyers really ask, keep them stable, and run them on a schedule, so a change in results means something. Second, add a free-text 'how did you hear about us' field to your demo form. It's crude, and it catches the assistant-influenced visits no tag can see.

If you want AI-search visibility alongside paid and owned media in one place, MediaPilot is Cresia's product for media and AI-search visibility. For the reporting side of the job, analytics teams is the relevant starting point.

Expect noise. Small volumes mean a handful of sessions can swing a weekly chart. Judge the trend over a quarter, and don't rebuild strategy around one week.

What not to do with Gemini attribution data

A few reflexes are worth resisting.

Don't treat the tag as a ranking lever. Nothing about adding or reading a UTM parameter makes your content more likely to be cited.
Don't build UTM-based links to your own pages and hand them out as Gemini. Faking the source pollutes the one clean signal you just gained.
Don't compare Gemini sessions to organic sessions on conversion rate without context. The audiences, intents and volumes differ, and small samples mislead.
Don't drop your other AI referrers. Other assistants may label traffic differently or not at all. A Gemini rule covers Gemini only.
Don't chase every new parameter. If a vendor changes values next quarter, you'll want one documented rule to update, not twelve dashboards.
Don't cut content investment because the referral line looks small. The click count is the smallest part of what answer engines do for a brand.

Where this leaves AEO and GEO programmes

The real shift is cultural. Teams doing answer-engine work have spent a year defending a discipline they couldn't measure properly. A labelled traffic source doesn't solve that, but it moves one piece from guesswork to evidence, and that's enough to change the conversation with finance.

The work that earns citations hasn't changed: pages that answer one question per section, consistent entity facts, working markup, and crawlable content. What's changed is that you can now see part of the result. Do the housekeeping, trend the numbers quietly for a quarter, and keep the visibility and brand lines next to it so nobody mistakes a clean click count for the whole story.

Frequently asked questions

Do Gemini's UTM parameters improve how often my site is cited?

No. They are added to links after a source has been chosen, so they affect what you can measure, not what Gemini retrieves or quotes. Citation depends on whether your passages answer the question clearly and can be crawled.

Will the new parameters replace referrer-based tracking?

Not entirely. A tag on the link survives where a referrer header may be stripped, so it fills gaps, but you'll still want your existing referrer rules for other AI platforms and for any Gemini surfaces that don't tag links. Check Google's documentation for the current scope.

How do I know my Gemini attribution is working?

Click through from a real Gemini answer to your site and watch the session appear in your analytics with the expected source. Then check that redirects, consent tooling and cache rules haven't dropped the query string on the way.

Is Gemini traffic big enough to matter yet?

That depends on your site, and your own data is the only reliable answer. For many sites it's small today, so judge it by trend and conversion quality, and keep measuring citations and brand outcomes alongside it.

Should I add UTM parameters to links I place in AI answers or profiles?

Only on links you control and publish yourself, such as a profile or a post, and keep your naming consistent with your tracking specification. Don't tag anything in a way that imitates Gemini's own source labels.

Sources

https://www.searchenginejournal.com/google-gemini-adds-utm-parameters-for-referral-attribution/591754/
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/