// HOW-TO · ANALYTICS

How to track Gemini traffic with UTM tags in GA4 (2026)

Track Gemini traffic in GA4 using its new UTM tags: read the AI Assistants channel, separate Gemini from AI Overviews, and avoid the self-tagging mistake.

Last verified · 2026-10-04 · by Moe Ameen

For a long time there was no clean way to see how much traffic Google's Gemini actually sent your site. Referrer data from AI assistants is unreliable — a large share of Gemini clicks, especially from mobile apps and Android assistant invocations, arrive with the referrer stripped and land in analytics as anonymous "direct" visits — so the channel looked far smaller than it was. In late September 2026 that started to change: Gemini quietly began appending UTM parameters to the links it surfaces, and Google's John Mueller confirmed he sees them too. A tag that rides inside the URL survives where a referrer doesn't, which finally gives you a sturdier signal to attribute.

This is the practical setup for reading that signal in Google Analytics 4. You will confirm the tag, lean on GA4's native AI Assistants channel for the referrer-detected portion, capture the tagged visits that referrers miss, keep Gemini separate from AI Overviews so the numbers mean something, and build one view you actually check weekly. Two things to hold onto before you start: the behavior is undocumented, so treat a tagged count as a floor rather than a complete total, and you never tag your own URLs to "help." For the why behind all of this — how the AI Assistants channel works and why referrers lie — see the guide on [tracking Gemini and AI-assistant traffic](/guides/tracking-gemini-ai-assistant-traffic).

The steps

Confirm Gemini is actually tagging your links. Before building any report, verify the behavior exists for you. Ask Gemini a question your site answers, click through on a cited link, and look at the landing URL in the address bar for appended UTM parameters identifying Gemini as the source. Because the change is undocumented and Google hasn't published the exact format or the conditions that trigger it, treat what you find as an early signal, not a contract — the tag can change without notice, so your reporting should degrade gracefully if it does.
Start with GA4's native AI Assistants channel — don't over-build. This is the step people over-engineer. GA4 already ships an AI Assistants channel that recognizes Gemini by its referrer (commonly gemini.google.com) and files those sessions automatically — no custom channel group or regex needed for the referrer-detected portion. Open Reports > Acquisition > Traffic acquisition, set the primary dimension to Session source / medium, and look for the Gemini rows. That is your baseline before any UTM work, and for many sites it is most of the picture.
Capture the UTM-tagged visits the referrer misses. The tag earns its keep precisely on the clicks that arrive with no referrer — the mobile and Android visits that would otherwise fall into "direct." In an Explore report, add Session source / medium or the landing-page-plus-query-string dimension and look for the Gemini UTM value in URLs that didn't come through as a clean referral. These are the sessions the native channel can't see, so adding them to your Gemini total is the whole point of the tag existing.
Keep Gemini separate from AI Overviews. This is the error most likely to make your AI reporting meaningless. Gemini the assistant (a click from a Gemini answer, which lands in GA4 as a session) and AI Overviews (the AI summaries atop Search results) are different surfaces that behave differently — counting them together produces a number that means nothing. Read Gemini-assistant clicks in GA4; read AI Overviews and AI Mode impressions in Search Console's generative-AI report. See [how to track Google AI Mode traffic in Search Console](/how-to/track-google-ai-mode-traffic-in-search-console) for that side.
Never tag your own canonical URLs with utm_source=gemini. When people learn Gemini tags its links, the tempting next move is to add utm_source=gemini to their own URLs to "track it properly." Do not. Gemini links to your real canonical URL, so self-tagging stamps that parameter onto every visitor — organic, email, social — mislabeling all of them as Gemini traffic and destroying the exact number you were trying to measure. Let Gemini apply its own tag; your job is to read it, not reproduce it.
Build one Gemini view you read weekly, and treat it as a floor. Turn the above into a standing exploration: source/medium and landing page as rows, sessions and your key events as columns, filtered to Gemini and the AI Assistants channel, saved so you glance at it each week rather than rebuild it. Label it clearly as a floor, not a full count, because the tag is undocumented and not every Gemini click will carry it. The value is the trend over a stable view — is Gemini growing as a referral source, and which pages earn the clicks.
Act on the number: the tag measures the click, it doesn’t earn it. Once the view confirms Gemini is a real referral source, remember what it can and can't do — it is a ruler, not a growth lever. The way you get more Gemini clicks is to be the source Gemini keeps citing, which is a content-and-consistency problem: publishing clear, verified, on-brand material across the surfaces the engines read, often enough to be a stable entity they trust. Feed the view's winning pages back into what you produce next, the same loop you'd run for any AI-visibility program.

Common gotchas

Treating the tagged count as complete. The behavior is undocumented and not every Gemini click carries the tag, so a tagged total is a floor — don't report it as the full number or build permanent reporting that assumes every click is tagged.
Conflating Gemini with AI Overviews. The assistant and the Search summary are different surfaces; merging their numbers produces a figure that means nothing. Read Gemini clicks in GA4 and AI Overviews impressions in Search Console.
Self-tagging your own URLs with utm_source=gemini. Gemini links to your real URL, so tagging it yourself mislabels organic, email, and social visitors as Gemini traffic and corrupts the metric.
Relying on referrers alone. A large share of Gemini clicks arrive with the referrer stripped (notably mobile and Android), so referrer-only detection undercounts the channel badly — the UTM capture is what rescues those sessions.
Judging the channel on last-click only. An AI assistant often opens a relationship a later search or direct visit closes, so check conversion paths before concluding Gemini drives nothing.
Expecting the tag to grow traffic. It measures clicks; it doesn't create them. Mistaking the measurement for the lever is how teams optimize the dashboard instead of the content that earns the citation.

Where Kompozy fits

GA4 is the read side of this: once the Gemini view is live, it tells you — with a signal that finally survives the referrer-stripping — that Gemini is a real referral source and which of your pages earn the clicks. What it can't do is act on that answer, and the answer is almost always the same one. Gemini clicks come from being the source Gemini keeps citing, and being cited is a footprint-and-cadence problem: the engine grounds its answers across many surfaces, so it rewards a brand that shows up consistently, in one voice, across the places it reads — not a site that published one good page and went quiet. Producing at that footprint by hand is where most creators stall, and it is exactly the write side [Kompozy](/) is built to carry.

Here is the loop the tag makes legible. Your Gemini view names the winners — a topic, a format, the pages pulling AI referrals — and you feed that back into what Kompozy generates: [blog articles and email newsletters](/glossary/output-buckets) an assistant can ground an answer in, platform-native text, [Persona Shorts](/glossary/persona-shorts) and avatar video, brand-exact carousels, and quote cards, all governed by one [Persona Brief](/glossary/persona-brief) so your name, claims, and voice stay identical across every surface — which is the consistency that makes you a stable entity an engine trusts enough to cite. Then [Autopilot](/glossary/autopilot) schedules the batch and fans it across the eight social platforms plus blog and email behind a per-post review gate, so the footprint Gemini reads keeps growing on a deliberate cadence instead of in bursts.

The honest boundary matters: Kompozy is not an analytics tool and will not read Gemini's tag for you — you keep GA4 for measurement, and nothing replaces building the UTM view yourself. What Kompozy changes is the input to that measurement: it turns the lever the tag points at — consistent, citable, multi-surface publishing — from a manual grind into a standing capability. Starter is $199/mo (5,500 credits) for a solo creator building that footprint; Pro is $499/mo (18,000 credits) for a team publishing daily across every surface; Enterprise is custom. The tag tells you Gemini is sending clicks; Kompozy is how you earn more of them.

Frequently asked questions

Does Gemini add UTM tags to its links?

Yes. In late September 2026 a Reddit user spotted Gemini appending UTM parameters to the links it surfaces, and Google's John Mueller acknowledged publicly that he sees them too. The working assumption is that the tag fires when Gemini cites or links a source in an answer. Google hasn't announced the change or published the exact format, so treat it as a real but unofficial signal — a floor on your Gemini traffic, not a complete count.

How do I see Gemini traffic in GA4?

Mostly you don't have to build anything for the referrer-detected portion: GA4 ships a native AI Assistants channel that recognizes Gemini by its referrer and files those sessions automatically. Open Reports > Acquisition > Traffic acquisition, set the dimension to Session source / medium, and look for the Gemini rows. Then add an Explore view to capture the UTM-tagged visits that arrive with no referrer, which the native channel can't see on its own.

Why does Gemini traffic show up as “direct” in my analytics?

Because the referrer is often stripped. A large share of Gemini clicks — especially from mobile apps and Android assistant invocations — arrive with no referrer header, so analytics files them as anonymous direct visits. That is exactly the gap the UTM tag closes: a parameter inside the URL survives the trip where a referrer doesn't, so the tagged clicks can be attributed to Gemini instead of disappearing into direct.

Should I add utm_source=gemini to my own URLs?

No. Gemini links to your real canonical URL, so adding utm_source=gemini yourself would stamp the tag onto visitors arriving from organic search, email, and social — mislabeling all of them as Gemini traffic and corrupting the number you're trying to measure. Let Gemini apply its own tag and read it in analytics; never apply it manually.

Is tracking Gemini the same as tracking Google AI Overviews?

No, and conflating them makes the data meaningless. Gemini is the assistant, and a click from a Gemini answer lands in GA4 as a session you can attribute. AI Overviews are the AI summaries at the top of Search results, which are largely zero-click; you read their impressions in Search Console's generative-AI report, not in GA4. Track the two separately.