Quick answer
Some visits from AI assistants carry a referrer such as chatgpt.com or perplexity.ai, and can be grouped in Google Analytics with a custom channel. Many arrive with no referrer at all, so analytics under-counts AI influence. Combine referral data with prompt tracking and server logs.
Key takeaways
- Group known AI referrers into one channel for reporting.
- Expect under-counting, because many AI visits have no referrer.
- Pair traffic data with observed mention and citation rates.
How do you set it up?
- 1In Google Analytics 4, open the traffic acquisition report and filter by source to see which AI domains already send visits.
- 2Create a custom channel group named AI assistants that matches those source domains using a regular expression.
- 3Place the AI channel above the default Referral channel so visits are classified correctly.
- 4Add UTM parameters to links you control, such as links in your own documentation or shared prompts.
- 5Review landing pages: the pages AI sends people to show what answers are citing.
What will analytics miss?
- Visits from apps that strip the referrer.
- People who read a brand mention in an answer, then search your name later.
- Answers that name you without linking at all.
Report the limits
Setting up the AI channel in GA4
- 1Open the Traffic acquisition report and look at Session source to see which AI domains already send visits.
- 2Create a custom channel group named AI assistants.
- 3Add a rule that matches the source domains you find, using a regular expression that lists them.
- 4Place the AI assistants channel above the default Referral channel so visits are classified there.
- 5Save, then check the Landing page report for the AI channel to see which pages assistants send people to.
Start with the domains you actually observe in your own data, and add others as they appear. The list changes as products and domains change, so revisit it every quarter.
What analytics will and will not tell you
| Situation | Shows up in analytics? |
|---|---|
| A person clicks a citation link and the referrer is passed | Yes, as a referral from the assistant's domain |
| A person clicks from an app that strips the referrer | Often as direct traffic |
| A person reads your name in an answer and searches for it later | As branded search, not as an AI visit |
| An answer names you without a link | No |
Because of this, AI referral traffic is best treated as a lower bound on AI influence. Report it as such, and pair it with mention and citation rates from your prompt set so leadership sees both influence and clicks.
Use landing pages as evidence
The pages assistants send visitors to reveal what answers are citing. A pricing page, a comparison or a documentation article showing up repeatedly tells you which content is doing the work. Compare those pages with your prompt results and invest in the ones already earning attention.
Server logs add a second view
- Logs show requests from AI crawlers and user-initiated fetches, identified by user agent, before analytics scripts run.
- They let you confirm which bots visit, how often, and which pages they read.
- Comparing them with the user agents each provider documents helps you confirm a crawler is who it says it is.
What not to conclude
A rise in AI referrals is encouraging, but it does not prove the effect of a specific change. Many things move at once. Keep a dated log of your changes and look for consistent patterns over several periods before claiming a link.
Which domains should I include in the channel?
Start with the AI domains that appear in your own source report, such as chatgpt.com and perplexity.ai, and add others as you see them.
Can server logs help?
Yes. Logs show crawler activity and referrers before analytics scripts run, which helps confirm which AI bots visit.
Is a rise in branded search a signal?
It can be, but it has other causes. Treat it as supporting evidence, not proof.
Which domains should I include in the channel?
Those you see in your own acquisition report, such as the domains of the assistants that already send visits. Add new ones as you notice them.
Can I tag links to measure better?
Yes, for links you control, such as those in your documentation or in content you share, add UTM parameters. Links inside AI answers are outside your control.
Cite this page
BrandWater AI Research. (2026, 7 September 2026). How to measure AI referral traffic in Google Analytics and your logs. https://brandwaterai.in/guides/measure-ai-referral-traffic
How we work
Figures are dated and linked to their sources. Where none exist we say so. Read our methodology and AI transparency pages.