Does Gemini Add UTM Tags to Links? What AI Attribution Still Misses

Google appears to have added UTM parameters to links in Gemini answers. A Reddit user spotted it and Google’s John Mueller replied that he saw it too. Google has not documented the change. UTMs show which visits came from Gemini. An AI brand monitoring tool shows what Gemini said before the visit. A client needs both.

What Was Spotted

Search Engine Journal reported the change on October 1, 2026. A Reddit user noticed that outgoing links from Gemini carried UTM parameters. The post described the change as very new, about a day old. Mueller answered: “Nice. I see it too.”

The article’s subheading says Google confirms the change. The support for that is Mueller’s reply. The article also says the implementation is undocumented. Treat this as an observed change that a Google employee acknowledged, not as a launch. Nothing in the report says how long the tags will stay or whether they apply everywhere. Do not treat them as a stable data feed until Google documents them.

Why Gemini Traffic Was Hard to Attribute

The report explains the problem. When a visit arrives with no referrer and no UTM data, analytics falls back to direct. Gemini visits can end up in GA4’s direct bucket this way. An agency sees a visit, but not where it came from.

That is a problem for reporting. A client asks whether AI assistants send any traffic. The honest answer used to be “some, but we cannot separate it.” Clean tags would let an agency compare Gemini visits with other AI sources and with traditional search.

The last row deserves attention. A buyer can read a Gemini answer, remember the client’s name and search for it later. That buyer may never produce a tagged visit. A tag cannot count a click that never happened.

What UTMs Can and Cannot Do

UTM tags answer one question well: which source sent this visit? They cannot answer the questions a client asks first.

This is the gap an AI brand monitoring tool fills. GA4 counts what arrived. A reading of the answer records what the assistant said to the buyer who may or may not have clicked.

What About ChatGPT?

The same article makes a point about ChatGPT. It says ChatGPT adds UTMs only to links in answers grounded in web sources, not to answers drawn from training data. The detail matters for attribution. Two assistants can both send traffic, and each can tag it under different conditions.

Do not build a report on the assumption that tagging is complete. The article itself says it is unclear when Gemini adds the tags, so the data’s completeness is unknown.

How to Check Your Own Reports

Google has not published the tag values, so do not guess them. Check what your own data shows.

1. Open GA4 and compare traffic sources before and after late September 2026. Look for new source or medium values.

2. Click a link from a Gemini answer yourself. Watch what appears in the address bar and in real-time reports.

3. Write down the exact values you see and the date you first saw them. Keep that note with the report.

4. Compare direct traffic before and after. A fall in direct traffic with a new named source may be related. Do not call it proof.

5. Re-check monthly. The behaviour is undocumented and may change without notice.

This routine gives you a defensible record. It says what you saw, when you saw it and what you did not know. It is also the base for AI visibility tracking across several assistants, because each assistant tags its links differently.

Mueller Asked for Referrers Too

One more detail in the report is easy to miss. Mueller said he would like referrer data kept alongside the UTM tags: “It would be great to have these retained (in addition to the UTM-tagging).” That is a wish from a Google employee, not a commitment. It shows the setup is not final.

For agencies, the lesson is practical. Do not hard-code a report around today’s tags. Keep the logic flexible and record which version of the behaviour each month’s data came from.

Mistakes Agencies Will Make With the New Tags

A new data source invites quick conclusions. Four habits are likely to cause trouble.

The second row needs one more line. A visit that used to land in direct and now carries a tag is the same visit with a new label. While the change settles, report direct traffic and tagged AI traffic together. Then a rise in one and a fall in the other will not look like growth.

Why Not Wait for Google to Document It?

Clients ask about AI traffic now. A dated note of what you saw is better than a pause. Write down the observation, the date and the open questions. When Google documents the behaviour, update the note. If Google changes it first, the note shows when your data shifted.

What an AI Brand Monitoring Tool Must Add

Visit counts are half of the story. A good tool supplies the other half. Five requirements matter.

SKYA publishes how every number is calculated. For the idea behind reading several assistants at once, see why five AIs judge your client.

A Report That Joins Visits and Answers

Together these sections turn a vague traffic line into a story. The client sees the answer, the visit and the fix on one page. That is the kind of AI visibility tracking that survives a hard question.

Questions a Client Will Ask

Plain answers build trust. A client who hears “we do not know yet, here is how we will find out” trusts the next number more.

What to Tell Clients

Frequently Asked Questions

Does Gemini add UTM parameters to links?

It appears to. Search Engine Journal reported on October 1, 2026 that outgoing Gemini links carried UTM tags, and Google’s John Mueller said he saw it too. Google has not documented the change.

Which UTM values does Gemini use?

The report does not name them. Check your own GA4 data for new source or medium values, and record the date you first see them.

Where did Gemini traffic show up in GA4 before?

The report says visits with no referrer or UTM data can fall into the direct bucket. That makes Gemini visits hard to separate from other direct traffic.

Can UTM tags show what Gemini said about my client?

No. A tag identifies the source of a visit. To read the answer and check it for accuracy, an agency needs stored readings of the assistant.

What is an AI brand monitoring tool?

An AI brand monitoring tool reads what AI assistants say about a client and records whether it is named, cited and accurate. SKYA does this daily across nine assistants and stores the answer behind every number.