Gemini UTM Parameters: What Marketers Can and Can't Measure Yet

What happened

Google Gemini has started appending UTM parameters to the links it sends users to. That gives publishers and marketers a new way to identify traffic coming from Google's AI assistant 12. The change surfaced the way many Google changes do. A Reddit user noticed the tags, and Google's John Mueller replied in the thread that he could see them as well 2. Search Engine Journal reports that Google has not formally documented the change 3.

That lack of documentation is the main story. Everyone agrees the tags exist. Nobody outside Google knows when they are applied, to which links, or on which surfaces 23. So the update is useful but uneven. Any early traffic numbers built on it should be treated with caution.

Why a tag matters when referrers already exist

AI referral traffic is hard to measure because the referrer header often goes missing. When Google Analytics 4 receives a visit with no referrer and no UTM, it files that visit under Direct 2. For sites trying to understand how much traffic AI assistants send, that bucket hides the answer.

One participant in the Reddit thread, posting as CrawlWarden, offered rough figures on how badly Gemini traffic loses its referrer:

• The referrer mostly survives only on desktop web.
• Only around 9 percent of visits from the Gemini mobile app pass it through.
• Android assistant invocations strip it completely.

These numbers come from a single user, not from Google or a systematic study 2. Mueller asked for something reproducible and offered to pass it to the team 2. That suggests Google is at least interested in the attribution gap, though it has made no commitment.

A UTM parameter sidesteps the problem. It travels in the URL, so it survives in-app webviews and appears in raw server logs even when the referrer is gone 2. ZeroClick Labs makes a similar point: the tags preserve source information when browser referrer data is missing 1. GA4 already has an AI Assistant channel, but the Gemini UTMs add a second, independent signal 1. Search Engine Journal likewise says the tags can identify visits regardless of referrer data 3.

Where the accounts diverge

The sources agree on the basics but differ in emphasis.

• ZeroClick Labs is the most upbeat. It frames the change as making AI referral traffic "easier to measure" and says independent testing suggests the rollout may be broader than first reports implied 1.
• IndexLab is more cautious. Because Google hasn't said which links get tagged, it argues early numbers should be read as a floor, not a full count 2.
• Search Engine Journal sticks to the facts: the tags exist, they help, and the timing of when they are added is undocumented 3.

These views can both be true. A rollout can be broader than first thought and still be inconsistent across apps, devices, and link types. Without documentation, one site may see heavy tagging while another sees little. Comparing datasets across organizations, or even across months, becomes risky.

ZeroClick Labs adds one boundary worth stressing. The update improves measurement only. It does not change rankings or whether a site is eligible to be cited by Gemini 1. Better tracking won't produce more Gemini traffic. It only makes existing traffic easier to see.

The broader context

The Gemini change arrived in a busy week for Google measurement and AI policy. Search Engine Journal's roundup also noted:

• Google's guidance on AI-generated content now tells sites to fact-check body copy and metadata by hand before publishing.
• A Q2 click-through-rate comparison from AWR spans the end of a Search Console impressions error, which may create a break in the data.
• There were updates on AI Mode monitoring eligibility and two AI Overviews lawsuits that a judge dismissed 3.

The Search Console issue is a reminder that the measurement problem goes beyond AI assistants. Search data itself can shift in ways that distort period-over-period comparisons 3. Undocumented UTM behavior in Gemini adds another variable that analysts must account for.

The takeaway

The practical case for these tags is strong. Mobile and assistant traffic has been leaking into Direct, and a URL-level tag is a sturdier signal than a referrer header that many environments discard 2. Sites should start segmenting Gemini-tagged sessions in GA4 and their server logs now.

What they shouldn't do is treat those numbers as complete or stable. Until Google documents when tagging applies, Gemini UTM counts are best read as a minimum estimate of AI-assistant traffic 2. They are useful for spotting trends within one property and unreliable for benchmarking against others. The likeliest outcome is that the signal improves as the rollout matures. For now, the gap between "the tags exist" and "the tags are consistent" is where analysts need to be careful.