Gemini tagged its outbound links with utm_source=gemini for 11 days in September, according to Cedric Huesler, and that let him measure real Gemini clicks at 2.1× what the referrer shows. The figure comes from AEM (Adobe Experience Manager) Operational Telemetry: real-user visits on 1,000+ domains, bots excluded. Huesler, who works at Adobe, published it on LinkedIn on October 16 as a personal post, not as an Adobe announcement.
What the tagged window showed
Visits with a gemini.google.com referrer stayed flat while the tag was on. Next to them, Huesler saw an equal number of tagged visits that arrived with no referrer at all. Of those, 76% were on mobile. They engaged like search visitors and landed on deep pages, so a normal setup counts them as “direct”.
Not every Gemini link carried the tag: about a quarter of referred visits arrived without it. Huesler applied that same gap, by device, to the visits with no referrer, and that is where his estimate of ~2.5× comes from. So the two numbers are different kinds of claims: 2.1× is measured, ~2.5× is estimated. He says the tag stopped on September 29.
Two limits on reading this. The data is AEM telemetry, which is not GA4 data and does not cover the web as a whole. And we found no Google documentation for the tagging, its start or its stop. Vizzybl, an AI-visibility vendor that monitors Gemini answers daily, wrote that UTMs started appearing gradually on September 10 and that 100% of the 4,235 links it captured on September 28 were tagged. It also says the tags “temporarily disappeared” after September 29. We have no evidence either way on whether they came back.
How much Gemini traffic lands in direct?
In Huesler’s data, about 41% of Gemini’s clicks sat in direct and were recovered by the tag, and about 18% more sat in direct without a tag, which is an estimate. Only 40% showed up attributed to Gemini. The rest is a small slice under Google or other referrers.
| Share of Gemini clicks | Where it lands |
|---|---|
| 40% | attributed to Gemini |
| ~41% | in direct, recovered by the tag |
| ~2% | under Google or other referrers |
| ~18% | in direct, untagged (estimated, mostly mobile) |
Huesler’s shares are rounded, so they add up to about 101%.
What GA4’s AI Assistant rule keys on
Google’s Help page for the GA4 default channel group says AI Assistant is the channel for users who arrive “from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok”. The rule table assigns it when “the medium exactly matches ‘ai-assistant’”, or when the medium is set to “ai-assistant” and the campaign to “(ai-assistant)” “if the referrer matches a list of AI Assistants”. Direct, in the same page, is for users who arrive “via a saved link or by entering your URL”.
Our reading: the channel keys on a referrer or on that exact medium. The page does not say what GA4 does with utm_source=gemini when no medium is set, and no source we read gives a medium for Gemini’s links, so we do not say which channel the tagged visits fell into. We covered the channel itself in our report on GA4’s AI Assistant default channel.
Why would the referrer vanish? Kenichi Suzuki, summarizing the thread on his blog, reports that Huesler answered John Mueller by saying native mobile and desktop apps usually don’t set an HTTP referrer unless developers do so explicitly, and that UTM parameters are easier and less error-prone. That fits the general mechanism: a UTM parameter in the link URL travels with the click, so it survives when an app drops the referrer header. It does not recover Gemini traffic that arrives untagged. Vizzybl also notes that Gemini links often carry #:~:text= fragments, which browsers never send to the server or to GA4.
What to do with your own numbers
Huesler’s advice for SEO and GEO work: use 2× your Gemini referral sessions as a floor, and ~2.5× as a working estimate. Look for the rest in direct, among mobile sessions landing on deep pages. To get your own ratio, compare sessions tagged utm_source=gemini that have no referrer against gemini.google.com referrals for September 18–28.
If you run it in GA4, for example in an exploration, keep his condition that the tagged sessions arrived with no referrer. Vizzybl also lists utm_source=geminiforbusiness for enterprise and Workspace links, so check for it too. The window is short and the sample is AEM sites, so your ratio may land elsewhere. His 2× floor comes from AEM telemetry; it is not a correction factor proven for every property.
