How to track ChatGPT and Perplexity traffic in GA4

Can Google Analytics 4 measure traffic from AI assistants?

Yes, when the click carries a referrer that GA4 recognizes. A visitor who follows a link from an assistant may appear under a source associated with that service.

Common sources to inspect include:

• chatgpt.com
• perplexity.ai
• claude.ai
• gemini.google.com
• copilot.microsoft.com

These are examples, not a guaranteed exhaustive list. Platforms can change domains, redirect links, suppress referrer information or open links in ways that affect attribution.

Some genuine AI-origin visits will show as Direct or another source. GA4 also cannot tell you how often your brand was mentioned in AI answers that nobody clicked.

Step-by-step: find AI referral sessions

1. Open your GA4 property and go to Reports → Acquisition → Traffic acquisition.
2. Select the Session source / medium primary dimension.
3. Use the report filter or search for AI-related sources. For more complex matching, consider an Exploration with a filter for session source.
4. Add Landing page + query string as a secondary dimension to see which pages received those sessions.
5. Compare the last 28 complete days with the preceding 28 days.

Google documents this source-and-landing-page workflow in its Landing page report guide. Menu options and report permissions may vary by property.

A starting regular expression for a custom Exploration is:

.*(chatgpt|perplexity|claude|gemini|copilot).*

Apply that to the appropriate session-source dimension, not to page URLs or campaign names. Inspect the actual dimension values first, because an overly broad pattern can catch unrelated sources.

Read the results correctly

Suppose a report shows the following fictional example:

This indicates 183 attributed sessions from those listed sources in the selected window, not 183 unique people and not necessarily every AI-origin session. Key events only appear if you have configured the relevant GA4 events.

Go one level deeper: compare landing-page engagement and key-event rates against your other channels. A small number of highly relevant sessions may matter more than a larger number of unqualified visits.

What GA4 cannot tell you

• No citation counts: GA4 measures site activity, not whether an AI assistant cited a page in an answer.
• No full referrer coverage: Some visitors arrive with missing or altered referrer information.
• No reliable ‘AI market share’: Google organic search results and AI answers overlap in ways a basic referrer filter cannot disentangle.
• No guarantee of business value: A referral session is not automatically a customer, signup or sale.

If you need visibility into what people typed into Google Search, that is a separate dataset in Search Console. Learn why Search Console clicks differ from GA4 sessions.

A useful reporting template

Use one monthly table with source, sessions, engaged sessions, key events, top landing page and period-over-period change. Note when tracking configuration changed. Keep AI referrals separate from direct and organic search so the report remains auditable.

If you combine GA4 with Search Console, state which numbers come from which source. Do not ask the AI to infer conversions from Search Console alone.

Build a repeatable AI referral report in GA4

A one-off search in Traffic acquisition is useful, but a consistent report is better for measuring change. In Explore, create a free-form exploration and define the dimensions and metrics explicitly.

Suggested dimensions

• Session source / medium — the traffic source for the session; do not confuse this with First user source.
• Landing page + query string — which site URL received the attributed session.
• Date — for a daily trend, subject to your report’s granularity and retention settings.

Suggested metrics

• Sessions — volume of attributed visits.
• Engaged sessions — visits that met GA4’s engagement conditions.
• Key events — configured meaningful actions, when available.

Apply a filter to the session source dimension for known AI assistant domains. Use a suitable regex or an exact-source list after reviewing the values in your property. Save the Exploration and reuse the same filter each reporting period. If the interface restricts dimensions or metrics, use a compatible exploration tab rather than mixing session-scoped and event-scoped fields without checking their meaning.

For each monthly report, document the domain filter, chosen date window and any analytics configuration changes. A change in referrer labeling can look like growth even if real referral volume did not change.

Example: which AI landing pages are worth improving?

Imagine this fictional report:

The API article brings more AI traffic, but the pricing page may be more valuable per visit. Before changing either page, inspect which events count as key events: a checkout can be more meaningful than a scroll.

For GSC Magic, a sensible primary conversion would be a successful waitlist signup, provided that event is actually configured and measured. Without event instrumentation, the table would not support signup-rate claims.

This leads to a more useful question than “How much AI traffic did I get?” — “Which AI-referred landing pages introduce qualified visitors to my product?”

Tracking quality checks before sharing the numbers

1. Confirm the GA4 tag is installed on all relevant landing pages and respects your consent configuration.
2. Ensure the signup or purchase event fires after a successful action, not when a user merely clicks a button.
3. Check that internal cross-domain navigation does not create unwanted new referrers.
4. Review traffic-source values rather than assuming your regex catches every AI service.
5. Keep Direct traffic separate; do not attribute all unexplained Direct sessions to AI.
6. Monitor engaged sessions and key events alongside traffic counts.

If you are comparing SEO traffic and GA4 numbers, review why GSC clicks differ from GA4 sessions.

FAQ: AI referral analytics

Can I measure how many times ChatGPT recommended my brand?

Not from GA4 referrals. GA4 records activity on your site, not unclicked mentions within AI-generated answers.

Is AI referral traffic the same as Google AI Overview traffic?

No. Clicks from Google search experiences may be attributed as Google organic search, while referrals from stand-alone AI assistants can appear under their own domains. Do not merge these sources without a clearly defined methodology.

Why does a known ChatGPT visit appear as Direct?

Referrer information can be suppressed by an app, browser or navigation path. Analytics attribution cannot recover an origin it never received.

Is ‘AI traffic conversion rate’ useful?

Yes, if your conversion/key event is defined consistently and there are enough sessions to interpret the result. For tiny samples, show counts as well as rates and avoid overinterpreting changes.

Questions to ask an analytics-connected agent

These questions require GA4 access. Read-only Search Console data is not enough. For broader AI-assisted SEO analysis, explore GSC Magic and see the prompts library.