DataWorks

Gemini's New UTM Tags Don't Mean More Traffic. They Mean Less Guessing.

Google Gemini now adds UTM parameters to some outbound links. Here's how to track Gemini sessions in GA4, avoid attribution corruption, and interpret the data.

DemGen Daily Editorial · October 20, 2026

A Conductor benchmark found AI platforms account for about 1.08% of website traffic across industries, and ChatGPT drives 87.4% of those referrals. Google Gemini's share? Small, hard to measure, and until recently, frequently invisible in GA4 because it showed up as Direct.

That just changed, partially. Gemini started appending UTM parameters to some outbound links in early October 2026. The operative word is "some." Google hasn't documented when or why UTMs get added, which means this is a better signal than before, not a complete one. If you treat utm_source=gemini as a definitive count of Gemini influence, you'll undershoot. If you ignore it, you'll keep dumping AI-referred sessions into the Direct bucket and making budget decisions on incomplete data.

Why the Referrer Alone Wasn't Cutting It

Gemini technically had a referrer. On desktop web, it mostly survived. On mobile? A Reddit user named CrawlWarden put it bluntly: referrer pass-through from the Gemini mobile app ran around 9%, and Android assistant invocations stripped it entirely. That pushed a large share of Gemini traffic into Direct, where it sat alongside bookmarks, typed URLs, and every other session GA4 couldn't classify.

Google's John Mueller confirmed the UTM addition on Reddit and asked for reproducible cases of referrer loss, offering to forward them to the team. His response suggests even Google's internal teams don't have full visibility into how the referrer behaves across surfaces. A UTM survives in-app webviews and shows up in raw server logs, which is why this matters more than a minor tagging update.

The GA4 Setup That Actually Works

In GA4, you need to inspect Session source/medium, not just look for utm_source=gemini in a standard report. Gemini traffic can appear as "gemini," "gemini.google.com," or under GA4's newer "AI Assistants" channel classification. These are distinct signals, and conflating them will produce misleading numbers.

OneMetrik's guidance is worth internalizing: use Explorations with session-scoped dimensions (Session source, Session medium, Session campaign, Session default channel group) to analyze AI-origin sessions and their downstream outcomes. Standard acquisition reports won't give you this granularity. One GA4 quirk that trips people up: the dimensions "Landing page + query string" and "Page path + query string" don't include UTM values. To see the full URL with parameters intact, you need the "Page location" dimension.

The Misuse That Will Corrupt Your Data

Some teams, eager to track Gemini, will be tempted to append utm_source=gemini to their own canonical URLs or internal links. Don't. When a non-Gemini visitor hits a URL carrying that parameter (through a shared link, a cached page, an email forward), they inherit the Gemini attribution. Now your email or organic traffic looks like AI traffic in your reports, and you're making decisions on polluted data.

The utm.new team flagged this explicitly: adding Gemini UTMs to your own URLs can misattribute other channels as Gemini, degrading the very decision-making you're trying to improve. Let Gemini tag its own links. Your job is to receive and validate, not to manufacture the signal.

What This Doesn't Measure

Tagged clicks are one observable slice. They don't capture citations without clicks, brand searches triggered by a Gemini answer, or the multi-touch journeys typical in B2B SaaS where a prospect reads a Gemini citation, Googles your brand two days later, and converts through a paid ad. UTMs measure a click event, not influence. Roger Montti at Search Engine Journal framed this correctly: Gemini's UTM tagging is an attribution improvement, not a traffic increase.

A rise in utm_source=gemini sessions in your GA4 may reflect improved labeling of previously unattributed visits rather than real demand growth. If your exec team sees the number and assumes Gemini is a growing channel, you'll need to contextualize that data before it turns into a misguided budget request.

The QA Step Most Teams Will Skip

Because Google's implementation is undocumented, validate manually. Pull a sample of Gemini answers that link to your site. Click through. Record the destination URL. Confirm UTMs survive any redirects in your stack (especially if you use link shorteners, CDN redirects, or marketing automation landing pages that strip query strings). Then check what GA4 records as Session source, Session medium, and Session default channel group for those visits. If UTMs are getting dropped by a redirect, you'll never see them in GA4, and you'll conclude Gemini isn't sending traffic when it is. This is a 30-minute QA exercise that prevents weeks of bad reporting.

The bottom line: Gemini's UTMs are a useful, partial signal arriving without documentation or guarantees. Treat them as a minimum measurable baseline for Gemini-referred sessions, not a ceiling. The teams that build proper GA4 Explorations, validate redirect chains, and resist the urge to self-tag will get directional attribution they can actually defend. Everyone else will just have a new number to argue about.

Frequently asked questions

How can we reliably identify Gemini-driven sessions in GA4 if Gemini only adds UTMs to some links?

Use GA4 Explorations with session-scoped dimensions: Session source, Session medium, Session campaign, and Session default channel group. Look for values like 'gemini' or 'gemini.google.com' in Session source, and check GA4's AI Assistants channel classification separately. Treat tagged sessions as a minimum measurable baseline, not a definitive total.

Which GA4 dimension shows Gemini UTM values on landing pages?

Use the 'Page location' dimension. GA4's 'Landing page + query string' and 'Page path + query string' dimensions do not include UTM parameter values, so they won't surface Gemini tags even when they're present in the URL.

How do we prevent Gemini UTMs from polluting attribution across other channels?

Never append utm_source=gemini to your own canonical or public URLs. If a non-Gemini visitor hits a URL carrying that parameter (via a shared link or cached page), they inherit the Gemini attribution. Let Gemini tag its own outbound links; your job is to receive and validate the signal, not manufacture it.

Could a spike in Gemini-tagged sessions indicate real traffic growth?

Not necessarily. A rise in utm_source=gemini sessions may reflect improved labeling of visits that previously landed in Direct, rather than an actual increase in Gemini-driven demand. Contextualize the data by comparing total session trends and checking whether Direct traffic decreased proportionally.

What QA steps should we take to confirm Gemini UTMs are being captured correctly?

Pull a sample of Gemini answers linking to your site, click through, and record the destination URL. Confirm UTMs survive any redirects (CDN, link shorteners, marketing automation pages). Then verify the resulting Session source, Session medium, and channel group values in GA4 match expectations.