An analytics manager opens her weekly report and notices something odd. Direct traffic has crept up, and a lot of it is landing on deep blog posts and product guides, pages nobody types into a browser from memory. She has no campaign to credit, no new backlinks and no obvious explanation. Somewhere in that bucket, she suspects, are visitors who came from an AI assistant that cited her pages.
She’s probably right, and the problem has just become a little harder. At the end of September, Google’s Gemini stopped adding a tracking tag to the links it cites, and the change only came to light a month later. For anyone trying to measure how much business AI answers send them, that matters.
It matters because measurement drives decisions. When the marketing manager asks whether the content investment is working, or the finance team asks which channels deserve more budget, the answer comes from the reports. If a growing source of visitors is hidden inside a catch-all bucket, the reports quietly understate the value of the work that earned those visits. Content that’s being cited and recommended can look like it’s underperforming, simply because the credit lands in the wrong place.
The timing makes it more pressing. Search is changing fast, with AI answers taking a larger share of the way people research and compare. Click-through from those answers is often low, so every identifiable visit carries more weight, and losing the ability to identify some of them is a real cost. At the same time, nobody can wait for the platforms to sort it out: Google hasn’t explained the change, and the way AI assistants pass visitors to websites is still being worked out in public.
In this post we’ll walk through what happened, why AI traffic so often lands in the wrong place in your reports, how to avoid misreading the data, and a practical plan for measuring AI referrals better. We’ll finish with a checklist and answers to common questions.
What Happened to Gemini’s Tracking Tag
Around the beginning of September 2026, Gemini began adding a UTM parameter, utm_source=gemini, to the source links it cites in its answers. Google didn’t announce it. A Reddit user spotted the tags, and John Mueller of Google replied in the thread that he could see them too. Search Engine Journal reported the addition on October 1.
Then the tag vanished. Server logs published by search consultant Natzir Turrado show it disappearing overnight around September 28 or 29, after roughly three weeks. Turrado raised it on October 27, and Search Engine Roundtable reported on October 30 that it could no longer find the tag on any Gemini links. Google hasn’t commented on either the addition or the removal.
Why This Matters for Measurement
Gemini added utm_source=gemini to cited links for about three weeks in September, then removed it without notice, and Google hasn’t explained why.
Search Engine Roundtable reports that Google “seems to have stopped appending UTM parameters on Gemini links.”
The tag mattered because it was the clearest way to tell Gemini visitors apart. Without it, you depend on the referrer, the information a browser passes about where a visitor came from. On desktop and the web, Gemini visits still arrive with a gemini.google.com referrer, so they can be identified.
Mobile is the weak spot. One analysis reports that Gemini’s iOS app opens links in its own in-app browser and doesn’t pass a referrer, so those visits show up as Direct in Google Analytics 4. According to the same source, the tag had previously revealed between 16 and 65 extra hidden clicks for every 100 sessions identified through a standard referrer. That comes from a single report, so treat the range as indicative. The practical point stands: some AI visitors are invisible to the standard methods.
The Trap: Fake Trends in Your Data
There’s a second problem, and it’s easy to miss. If you looked at Gemini traffic in September, you’d have seen it appear, grow for about three weeks and then collapse. That pattern says nothing about interest in your content. It reflects a tag switching on and off.
ChatGPT had a similar wobble. According to reporting, ChatGPT links lost their utm_source=chatgpt.com tag between about October 7 and 16, before it returned, and the gap was later described as a bug. Anyone comparing October with September could easily read a drop in AI traffic that never happened.
The lesson is to annotate your reports. Mark the dates when sources changed how they tag links, and treat any sharp rise or fall that lines up with them as a measurement artefact until proven otherwise. We’ve made a similar point about not over-reading platform data in our post on Google adding links to AI answers, where Search Console reports impressions but not clicks.
Why “Direct” Keeps Growing
Direct traffic has always been a catch-all. It holds people who typed your address, bookmarked you or clicked from an app that doesn’t pass a referrer. AI assistants add a new source of exactly that kind of visit, particularly inside mobile apps.
That doesn’t mean all direct traffic is AI, and it doesn’t mean you should reassign it wholesale. It means a rising direct line, especially to informational pages that people rarely type in, deserves investigation. It’s a signal worth chasing, not a conclusion.
A Practical Plan for Measuring AI Referrals
None of these steps is perfect on its own, but together they give a much better picture than the default reports.
Create an AI channel in your analytics. In GA4, a custom channel group lets you gather known AI sources, such as chatgpt.com and gemini.google.com plus the equivalents you see in your own reports, into one place instead of scattering them across Referral and Direct. Check which sources actually appear in your data before you build it.
Look at the server logs. Logs record details that analytics tools miss, including the user agent. One analysis reports that Gemini’s iOS visits can be found by looking for a particular token in the user-agent string, which gives you a way to count visits that analytics labels Direct. Ask your developer or host whether this is feasible.
Watch landing pages, not just sources. Compare direct traffic to deep informational pages against your usual baseline. A jump on pages that people rarely type in is a useful clue, though only a clue.
Ask people. Add a “How did you hear about us?” field to enquiry and checkout forms, with options such as “An AI assistant.” Self-reported answers are imperfect, but they capture visits no tag can.
Track brand demand. If AI answers are mentioning you, branded search and direct visits to your home page often rise. Watching those alongside your content performance helps you see influence that clicks don’t show.
Tag what you control. For links you place yourself, in newsletters, profiles and partner pages, use your own UTM parameters. You can’t rely on others to tag them for you.
Combine it with search data. Google Search Console’s AI reports show impressions in AI Overviews and AI Mode, though not clicks. Together with your own analytics they give two partial views of the same activity.
Automate the reporting. A weekly view that brings these sources together saves hours and keeps the annotations consistent. That’s a natural fit for automation, and our digital marketing team builds this kind of reporting for clients.
What Not to Conclude
Be careful about two opposite mistakes. One is deciding AI traffic is negligible because the identifiable slice is small. Much of it may be sitting in Direct, so the true figure is probably higher than your reports show. The other is attributing every rise in direct traffic to AI. Without supporting evidence, that’s guesswork.
A better stance is to state what you can measure, name what you can’t and estimate the gap honestly. It’s the same discipline we recommend in our piece on why agentic AI makes AEO non-negotiable: visibility and traffic are different things, and both need tracking.
An AI Referral Measurement Checklist
- Annotated your reports with the dates Gemini and ChatGPT changed how they tag links
- Created an AI channel grouping in your analytics, based on sources you actually see
- Checked direct traffic to informational pages against your usual baseline
- Asked your developer or host about log analysis, including user-agent checks for mobile apps
- Added a “how did you hear about us” question with an AI assistant option
- Set up branded search and direct home page visits as supporting indicators
- Used your own UTM parameters on every link you place yourself
- Compared Search Console AI impressions with your own analytics
- Built a weekly report that combines these views and records known tracking changes
Frequently Asked Questions
Did Gemini stop adding UTM parameters? Yes, according to server-log data from consultant Natzir Turrado and a report from Search Engine Roundtable on October 30, 2026. Gemini added utm_source=gemini to cited links for roughly three weeks in September and then stopped. Google hasn’t commented.
Can I still track Gemini traffic? Partly. Desktop and web visits still arrive with a gemini.google.com referrer, but mobile app visits may appear as Direct because the app doesn’t pass a referrer.
Why does AI traffic show up as Direct? Some AI apps open links in their own browser without passing a referrer, and without a UTM tag, analytics tools have nothing to attribute the visit to, so they file it under Direct.
Did ChatGPT also lose its tracking tag? Reportedly, for about a week in October, between roughly October 7 and 16, before it returned, and the gap was later described as a bug.
Should I stop using UTM parameters? No. Keep using them on links you control. Just don’t depend on AI platforms to add them for you.
Where to Start This Week
If you only have time for a few actions, begin with these.
- Annotate your analytics. Add notes marking early September, when Gemini began tagging links, the end of September, when it stopped, and the week in October when ChatGPT’s tag went missing. It takes ten minutes and prevents months of misreading.
- Build one AI channel view. Create a custom channel group for the AI sources you actually see in your reports, so they’re counted together instead of scattered.
- Add one question to a form. Put “How did you hear about us?” on your main enquiry or checkout form, with an option for an AI assistant. Within a few weeks you’ll have a rough second opinion on your analytics.
- Ask for a log check. Ask your developer or host whether they can look at server logs for visits that analytics labels Direct, especially from mobile apps.
None of this needs new software or a big budget. It needs someone to own the question of where visitors really come from, which is often the missing piece.
Measure the Gap, Not Just the Number
The Gemini change is a small technical event, but it points to something bigger: the way AI assistants send people to your site is still being worked out, and the tools you use to measure it were built for a different era. The teams that cope best aren’t waiting for platforms to fix attribution. They’re building a measurement approach that holds up when tags appear and disappear, and being honest about what they can and can’t see.
It’s also worth remembering what the numbers are for. Attribution isn’t about winning an argument over credit between channels. It’s about knowing which work is paying off so you can do more of it. A report that says “we can see this much, and here’s roughly how much more we think is hiding” is more useful, and more honest, than one that pretends the tracking is complete. Tags will keep changing, apps will keep passing less information, and new assistants will keep arriving. A measurement habit built on several imperfect signals will outlast any single tag.
If you’d like help setting that up for your own site, our digital marketing team can build reporting that combines these views, and our automation work keeps it running every week without manual effort. You can also see how we approach reporting and measurement across other brands in our case studies.
