How to track AI referral traffic in GA4

AI search measurement has two different questions. Did an assistant mention or cite the brand? Did a person click through and convert? GA4 helps with the second question, while an AI visibility platform is needed for the first.

Keeping those layers separate prevents a common reporting mistake. A useful answer can influence a purchase without sending a click, and an AI referral can arrive after a branded mention that happened earlier in the journey.

Start with the GA4 Traffic acquisition report

Google defines Session source as the publisher or location that originated a session, and Session source / medium as the source and method associated with that session. Open Reports, then Acquisition, then Traffic acquisition, and inspect these dimensions rather than relying only on the default channel group.

The current GA4 Traffic acquisition documentation explains the available session-scoped fields. OpenAI also states that ChatGPT search referral URLs automatically include utm_source=chatgpt.com, which gives publishers a direct source to inspect. See the OpenAI publisher FAQ.

Build the source inventory from real data

Do not assume every assistant uses one permanent hostname. Start with the values already collected by your property, then maintain a reviewed list. Common candidates may include:

chatgpt.com
chat.openai.com
perplexity.ai
claude.ai
gemini.google.com
copilot.microsoft.com

Search each candidate in Session source and Session source / medium. Confirm that the traffic lands on plausible pages and is not a self-referral or spam source.

Create an AI referral comparison

In an Exploration or customized report, apply a filter to Session source using a reviewed regular expression such as:

^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com)$

Use the expression as a starting point. Adjust it to match the exact source values observed in your property. Compare at least these dimensions and metrics:

Break results down by landing page. AI referrals often arrive on detailed guides, comparisons, pricing, documentation or product pages rather than the homepage. Those entry pages show which content is valuable enough to cite and click.

Track business events, not visits alone

Mark the actions that represent real progress as GA4 key events: qualified form submission, booked demo, account created, checkout started or purchase completed. Avoid treating every scroll or button click as a conversion.

For a lead-generation site, a useful report can show:

Use consistent UTM parameters on links you control, such as campaigns or partner placements. You cannot add UTMs to links that an assistant creates independently, except where the platform itself supplies them.

Understand the direct-traffic blind spot

GA4 labels a session as direct / none when it has no clear referral source. Links opened through apps, copied into a browser, passed through redirects or filtered by privacy tools may lose source information. Google's direct traffic documentation describes these common causes.

That means measured AI referrals are a lower bound. Do not reclassify all direct traffic as AI traffic. Instead, use supporting evidence such as landing-page patterns, customer survey responses and answer-engine citations while keeping the attribution limitation explicit.

Add Search Console data for Google AI features

In June 2026, Google announced dedicated Generative AI performance reports in Search Console for AI Overviews, AI Mode and generative features in Discover. The initial rollout covered a subset of websites, so access may vary. Check the current property rather than assuming the report is available.

When present, use the report for impressions and clicks from Google's generative search surfaces. Use GA4 for post-click behavior. Read the Search Console announcement for the current scope and rollout notes.

Connect traffic with answer visibility

GA4 cannot tell you:

• how often an assistant mentioned the brand without a click
• which competitor was recommended instead
• which uncaptured source influenced an answer
• whether the brand description was positive or accurate
• how visibility differed by prompt and engine

Pair analytics with a fixed set of buying prompts. Track mention rate, citation rate and share of voice, then connect cited pages with GA4 landing pages and key events.

This combined view prevents both overclaiming and undercounting. A visibility gain with no traffic may still be early influence. Traffic with weak engagement may point to a poor landing-page match. Strong citations, qualified visits and conversions together are the clearest outcome.

Use a simple monthly reporting template

Report the following in one page:

1. AI mention and citation movement for the stable prompt sample
2. top cited domains and pages
3. AI referral sessions by verified source
4. landing pages and engagement
5. key events and value
6. changes shipped during the period
7. data limitations and the next action

SeoWave covers the answer-visibility side with multi-engine prompts, citations, competitors and trend evidence. Combine that with your GA4 and Search Console properties for a complete measurement loop. Start with how to measure AI visibility or run a scan.