How to Track ChatGPT Referral Traffic in GA4 (2026)

ChatGPT referral traffic can be tracked in GA4 by filtering session source for chatgpt.com and reviewing Google's AI Assistant channel. OpenAI now documents that ChatGPT referral URLs automatically include utm_source=chatgpt.com. Connect those sessions to landing pages and conversion events, but keep them separate from brand mentions and citations that produce no click.

What counts as ChatGPT referral traffic?

ChatGPT referral traffic is a website visit that begins when someone clicks a source or link inside ChatGPT. It is narrower than ChatGPT visibility: a page can be cited without receiving a click, and a brand can be mentioned without a linked source.

OpenAI's updated publisher documentation says ChatGPT automatically adds utm_source=chatgpt.com to referral URLs. It also tells publishers to allow OAI-SearchBot if they want pages eligible for summaries and snippets in ChatGPT search (OpenAI Publisher FAQ). These are related but separate signals: crawler access supports discovery, while the UTM parameter supports attribution after a person clicks.

Do not treat every request from an OpenAI user agent as a human visit. Bot requests, retrieval calls, previews, health checks, and server-side scans belong in logs, not audience analytics. A referral session requires a browser pageview or equivalent client-side event from a visitor.

Where does ChatGPT traffic appear in GA4?

Google Analytics 4 can expose ChatGPT visits through several dimensions. The most useful starting point is Session source, because GA4 maps UTM campaign values and referrer information into traffic-source dimensions during processing (Google Analytics traffic-source documentation).

Use this diagnostic table before building a dashboard:

GA4 view	Dimension to inspect	Expected signal	Best use
Traffic acquisition	Session source	chatgpt.com	Count sessions that started from ChatGPT
Traffic acquisition	Session default channel group	AI Assistant when classified	Compare AI traffic with search, social, and referral
Landing page report	Landing page + session source	Entry URL paired with chatgpt.com	Find pages that earn clicks
Events report	Event name + session source	Conversions from ChatGPT sessions	Measure checker use, leads, and sign-ups
Exploration	Session source + landing page + event	A custom funnel	Analyze the path from referral to outcome

Google's current default channel documentation includes an AI Assistant channel for sources such as ChatGPT, Gemini, DeepSeek, Copilot, and Grok. It explicitly excludes Google's own AI Overviews and AI Mode from that channel (GA4 default channel groups). This makes the channel useful for a high-level view, but use the source dimension when you need ChatGPT specifically.

OpenAI documents utm_source=chatgpt.com; it does not promise a fixed utm_medium or campaign name for every link. Build filters around the source rather than assuming a medium that may not exist. Check both chatgpt.com and any hostname variations already present in your own data before finalizing a rule.

How do you build a reliable GA4 report?

Start with an exploration so you can validate the raw classification before changing property-wide reporting.

1. Open Explore and create a free-form exploration.
2. Set the date range to at least 28 days unless the site already receives meaningful AI traffic.
3. Add Session source, Landing page + query string, Sessions, Engaged sessions, Key events, and Total users.
4. Add a session-scoped filter where Session source exactly matches or contains chatgpt.com.
5. Place landing page in rows and compare sessions, engagement, and key events.
6. Save the exploration with a clear name such as “ChatGPT referral traffic.”

Why use session scope? The question is which source began the visit. Event-scoped source can help with individual interactions, while first-user source answers who originally acquired the user across their recorded history. Mixing those scopes can produce reports that look inconsistent even when GA4 is behaving as designed.

If the AI Assistant channel is available and classifies the visits correctly, use it for executive reporting and retain the source-level exploration for diagnosis. If classification is incomplete, GA4 allows Editors to create a custom channel group with ordered, rule-based definitions (GA4 custom channel groups). Test the group before making it the property's primary channel because ordering determines which rule receives the traffic.

Which conversion events should you connect to ChatGPT?

Sessions alone do not show whether AI traffic creates value. Track the actions that correspond to the site's actual funnel.

For a SaaS or lead-generation product, a compact measurement plan might be:

Funnel stage	Event	When it fires
Qualified visit	page_view plus landing-page context	A ChatGPT session reaches a relevant page
Product interest	checker_start	The visitor starts the free analysis
Lead	generate_lead	The visitor submits and verifies a report request
Registration	sign_up	An account is successfully created
Activation	site_created or audit_start	The user adds a site or launches an audit

Google lists generate_lead, login, and sign_up among its recommended GA4 events. Recommended names improve consistency, but only send them after the action succeeds; a button click that ends in validation failure is not a completed sign-up (GA4 recommended events).

Attach only the parameters needed for analysis, such as plan, language, or content group. Do not send email addresses, names, API keys, or other personally identifiable data to analytics. Verify each event in DebugView or Realtime, then mark the business-critical events as key events.

The free AI visibility checker can be the first conversion step for visitors who arrive from an informational ChatGPT citation. A landing page should make that next action obvious without hiding the answer that earned the citation.

How do you track ChatGPT traffic in Plausible?

Plausible supports utm_source, utm_medium, utm_campaign, utm_content, and utm_term as acquisition parameters. Its documentation says those values are available in Campaigns while the parameters are stripped from the displayed page path (Plausible custom query parameters).

To isolate ChatGPT visits:

• open Campaigns and filter UTM Source for chatgpt.com;
• compare entry pages, visit duration, bounce rate, and conversions;
• define goals or custom events for the same successful funnel actions used in GA4;
• verify that the tracking script loads on public landing pages and that consent or browser controls are not preventing collection;
• compare totals directionally rather than expecting GA4 and Plausible to match exactly.

Plausible uses a privacy-first, last-click model for conversion attribution, while GA4 has different identity, session, consent, and attribution behavior. The tools can disagree because they are measuring with different rules, not necessarily because one is broken (Plausible traffic sources and campaigns).

Why might ChatGPT visits be missing?

Missing data has several possible causes. Diagnose the collection chain before concluding that ChatGPT sent no traffic.

No one clicked. A page may be cited or summarized without generating a visit. Citation visibility and referral traffic are different metrics.

The URL was copied or shared. A user may copy a destination into a new tab, a messaging app may strip attribution, or a privacy tool may remove parameters and referrer information. The visit can then appear as direct traffic.

The tag did not run. Content security policy, consent configuration, ad blockers, JavaScript errors, or a missing production measurement ID can prevent the pageview from reaching analytics.

The filter is too narrow. Exact-match rules can miss source variants. Inspect raw session sources before choosing a regular expression, and avoid a broad filter such as “contains ai,” which can mix unrelated sites.

The event fired too early. A conversion event attached to button press can count failed requests. Fire it from the success state or confirmed server response.

Reporting is delayed. Realtime and standard reports do not share identical processing. Use Realtime for installation checks and consolidated reports for trend analysis.

How should ChatGPT traffic be reported with GEO metrics?

Use three connected but distinct layers:

1. Eligibility: OAI-SearchBot access, successful responses, indexability, rendered content, and canonical URLs.
2. Observed visibility: prompt-level mentions, citations, cited pages, sentiment, and competitors.
3. Business impact: ChatGPT referral sessions, landing pages, checker completions, leads, sign-ups, and activation.

The three-layer AI visibility guide explains why one readiness score cannot prove a citation or conversion. Use the AI mention-tracking framework for prompt observations and the ChatGPT crawlability checklist for technical eligibility. The broader GEO hub connects those measurements to content and entity optimization.

A weekly report should show current and previous equivalent windows, identify incomplete dates, and list the landing pages that moved. Keep bot requests out of the audience count. If five visits produce two qualified sign-ups, that can matter more than hundreds of unqualified sessions, but a tiny sample should still be labeled as directional rather than statistically stable.

FAQ

Does ChatGPT always add utm_source=chatgpt.com?

OpenAI's publisher FAQ states that ChatGPT automatically includes utm_source=chatgpt.com in referral URLs. Attribution can still be lost if a parameter is stripped, a URL is copied, tracking is blocked, or the visit never triggers the analytics tag.

Is ChatGPT traffic classified as referral or AI Assistant in GA4?

GA4's current default channel definitions include an AI Assistant channel. Use Session default channel group for the category and Session source for the specific chatgpt.com source. Existing reports or custom groups may classify historical data differently.

Can GA4 show which ChatGPT prompt generated the visit?

Not reliably. The documented UTM value identifies ChatGPT as the source, not the user's full prompt. Do not invent prompt attribution from a landing page. Pair analytics with separate prompt and citation monitoring.

Why are GA4 and Plausible ChatGPT totals different?

They use different session, privacy, consent, identity, and attribution rules. Ad blockers and tag availability can also affect one system differently. Compare trends and verify implementation rather than forcing the totals to match.

Does zero ChatGPT referral traffic mean the site is invisible?

No. The site may be mentioned or cited without a click, or traffic may have lost attribution. Check crawler access, fixed prompt observations, citations, server delivery, and analytics collection before drawing a conclusion.

Sources

• OpenAI: Publishers and Developers FAQ
• Google Analytics: Campaigns and traffic sources
• Google Analytics: Default channel groups
• Google Analytics: Custom channel groups
• Google Analytics: Recommended events
• Plausible: Tracking custom query parameters
• Plausible: Traffic sources, referrers and UTM campaigns