How to Track ChatGPT and AI Traffic in GA4

Quick answer: GA4 now sorts visits from recognized AI assistants into a built-in AI Assistant channel. For a rule you control, build a custom channel group with an AI channel that matches assistant domains by regex, and place it above Referral. Both methods count only visits that arrive with source data, so unclicked citations and stripped referrers stay outside the total.

When Google showed an AI summary, users clicked a traditional result in 8 percent of visits, against 15 percent when no summary appeared. They clicked a link inside the summary in just 1 percent of visits. Most AI exposure never becomes a session GA4 can count.

The visits that do arrive run into three attribution gaps. Google’s AI Assistant channel doesn’t name Perplexity, links that reach you without a referrer land in Direct, and clicks from AI Overviews stay inside Organic Search.

Since May 13, 2026, GA4 has assigned visits from recognized AI assistants to an AI Assistant channel in the Default Channel Group. Start there, then add a custom channel group when you need a rule you can see, edit and audit. Treat either total as attributable AI visits, not all AI influence, and pair it with landing pages, key events and your visibility in AI answers.

AI Traffic Tracking Methods Compared

Check the built-in AI Assistant channel

1. See what GA4 already classifies

   Open the Traffic acquisition report and look for the AI Assistant row. When a referrer matches a recognized assistant, GA4 sets the medium to ai-assistant and the campaign to (ai-assistant).

   Google names ChatGPT, Gemini, Deepseek, Copilot and Grok in its channel definition, and its release note adds Claude. Neither the channel definition nor the release note lists Perplexity as an AI assistant.

   That gap is the main reason to build your own group. A custom rule puts the assistant list in your hands, and you can document every entry.

Create a custom channel group for LLM referral traffic

2. Open the channel group controls

   Go to Admin > Data display > Channel groups > Create new channel group. Give it a clear name, such as Acquisition with AI, so colleagues know its purpose.

   Google allows 2 custom channel groups in a standard property and 5 in an Analytics 360 property, on top of the default group. Each group holds up to 50 channels.

   Check existing groups before you create another. One shared definition keeps content, acquisition and leadership reports aligned on what counts as AI traffic.

Add an AI channel and source condition

3. Build the rule around the source

   Inside the new group, choose Add new channel > Add condition group. Name the channel AI Assistants, pick Source, and select matches regex.

   This follows Google’s own AI assistants example for custom channels. Add a second condition set to OR: Medium exactly matches ai-assistant. That catches assistants Google recognizes but your regex misses.

   Matching on Source also covers tagged links. When a ChatGPT link carries utm_source=chatgpt.com, GA4 reads chatgpt.com as the source, so the Source condition matches it whatever medium the visit records.

   Classify by source, not by topic. A page about ChatGPT doesn’t prove its visitors came from ChatGPT.

   Record the channel name and rule in your reporting notes. When another analyst reviews the same period, every change has an explanation.

Paste the AI referral regex

4. Match the specified assistant domains

   Use this exact pattern in the Source condition with matches regex:

   chatgpt\.com|chat\.openai\.com|perplexity\.ai|www\.perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|copilot\.com|claude\.ai|chat\.deepseek\.com|grok\.com

   Vertical bars separate the alternatives. Each backslash turns the next dot into a literal period, so every domain stays exact.

   The pattern covers ChatGPT and its legacy domain, Perplexity with and without www, Gemini, both Copilot domains, Claude, Deepseek and Grok. Keeping chat.openai.com in the rule keeps older periods comparable.

   GA4 regex is full match and case sensitive by default, and Google points to RE2 syntax as its reference. The whole source value has to match one alternative.

   Skip broad fragments like ai or gpt. Google’s example uses wide patterns, but exact domains are easier to audit and pull in fewer unrelated sites.

Place AI above broader channels

5. Set the matching priority

   Return to Admin > Data display > Channel groups and open your group. Move AI Assistants above Referral, and above Organic Search if your rules call for it.

   Google’s AI assistants example calls for this order. GA4 assigns each visit to the first channel whose definition matches.

   Referral covers non-ad links from other sites and apps. A link from an assistant Google doesn’t recognize fits that bucket unless an earlier channel claims it.

   Order matters as much as the regex. A correct rule placed below Referral still loses its visits to Referral.

   Reordering changes how existing source data is sorted. It doesn’t restore missing referrers.

Validate sources and historical classification

6. Check the rule against observed values

   Open Explore and start from a blank canvas or a template. Explorations let you filter on dimensions, metrics or both.

   Filter Session source or First user source with matches regex and paste the same pattern. Session source shows what brought in each visit. First user source shows how a person first arrived, which matters when someone finds you through an assistant and returns later through search.

   Review the matching values before you treat the channel total as a baseline. If an expected source is missing, check the full value, spelling and case, and compare it with your alternatives before you widen the rule.

   Custom channel groups apply to reports retroactively, so review a past period as well as the current one. A source rule doesn’t depend on the ai-assistant medium, so it also sorts sessions from before GA4’s May 2026 change.

   Reclassification works only with the attribution data GA4 already holds. It won’t name the assistant behind a visit that arrived with no source.

Report AI acquisition and key events

7. Connect attributable visits to business outcomes

   Use your custom group in acquisition reports to compare the AI channel with every other channel. For a deeper cut, filter an Exploration by Session source with the same regex.

   Review AI visits, their landing pages and the key events (formerly conversions) and revenue metrics your property already measures. Keep the date range and key event definitions the same across channels.

   Compare the Traffic acquisition report with the User acquisition report, which uses First user source. The second view shows key events by the source that first brought each user in, so an assistant that introduced a buyer keeps its credit.

   A visit total alone won’t tell you whether those visits drove business. Tie every AI number to a landing page and an outcome.

   Keep citation monitoring separate from visit attribution. Our guide to tracking your brand in ChatGPT and Perplexity covers the visibility side.

   Report two answers clearly: where the brand appeared, and what identifiable visitors did after they arrived.

Why GA4 Undercounts AI Traffic

Citations don’t always produce visits

A citation in an AI answer and a session on your site are different events. A reader who takes the answer and never clicks the source leaves no trace in GA4, the zero-click pattern behind the Pew Research Center figures in the opening.

Pew measured Google searches by 900 U.S. adults in March 2025, not every assistant. The gap still shows why citations and referred visits need separate reporting.

Missing referrers leave attribution incomplete

When an AI citation opens without a referrer header or a utm_source=chatgpt.com parameter, GA4 has no evidence of the assistant that sent the visit. This dark AI traffic lands in Direct, and the AI Assistant channel misses it too.

In October 2026, Search Engine Roundtable reported ChatGPT links dropping both UTM tags and referrers. A follow-up update called it a bug, and the tags came back.

Keep unexplained Direct traffic separate from confirmed AI referrals. Labeling every unattributed visit as AI swaps missing evidence for a guess.

Return visits hide the first AI touch

When a person discovers a brand in an assistant and returns later through branded search, Session source credits Google for the return visit. First user source keeps the assistant as the acquiring source.

Check both scopes before you judge AI’s contribution to key events. A session-only view undervalues assistants that start the journey.

Copied answers show up as branded search

Readers who copy a brand name or URL from an unlinked AI answer into a new tab arrive as Organic Search or Direct, with no assistant in the source.

Compare branded query impressions in Search Console with Direct landings on the same pages to spot that pattern. Report it as a signal, not a confirmed AI referral.

Google AI clicks stay in Organic Search

Google includes AI Overviews and AI Mode in Organic Search and excludes them from the AI Assistant channel. Those clicks show up as ordinary Google organic traffic.

Search Console reports AI feature appearances in the Performance report under the Web search type. Google also added a dedicated generative AI view that sits alongside the overall report.

Moving your AI channel above Organic Search won’t split out Google AI clicks. Your source rule still needs a matching source value.

Common Tracking Mistakes

Most errors start with a definition that’s too broad or a total read as more than it is. Check these before you share an AI number:

• Channel order: Keep AI Assistants above Referral so the broader rule doesn’t claim matching visits first.
• Regex matching: Use full domains and inspect real source values before you change the rule.
• Medium-only rules: Pair the ai-assistant medium with a Source condition so assistants Google doesn’t recognize still match.
• Native vs. custom totals: Label which definition each report uses. Google’s AI Assistant channel and your custom rule cover different lists.
• Session-only views: Check First user source before you judge AI’s effect on key events.
• Missing attribution: Keep Direct separate unless source evidence names the assistant.
• Visibility vs. visits: Report citations and sessions as two different measures.
• Rule history: Log every rule change so colleagues know what the channel includes.

Pair acquisition reports with AI visibility analysis when you judge answer-engine presence. Referral data shows what visitors did after clicking. Visibility data shows whether the brand appeared.

Where AEO Agent X Fits

AEO Agent X Growth connects Google Search Console and GA4, at $449/mo. Our platform tracks positions, clicks, indexation, leads and AI citations per URL, so content decisions come from your own data.

We run SEO, AEO and GEO work approval first: nothing goes live until you say so. Feedback is stored as a rule and applied to everything written afterwards, so a correction stays corrected.

Native Channel or Custom Rule: Decide Before Your Next Report

Use the built-in AI Assistant channel for a quick read with zero setup. Build a custom channel group when you need assistants Google doesn’t name, such as Perplexity, sessions from before May 2026, or a documented rule you control.

Either way, start with a narrow rule, the right channel order and one key event definition. Before you share results, separate confirmed assistant referrals, Google organic traffic and visits with no source.

Request access to AEO Agent X for approval-first content work built on your own Search Console and Analytics data.

Frequently Asked Questions (FAQs)

How do I track ChatGPT traffic in Google Analytics 4?

Check the built-in AI Assistant channel in the Default Channel Group first. For full control, create a custom channel group with an AI channel that matches chatgpt.com and chat.openai.com by regex on Source, which also catches links tagged utm_source=chatgpt.com. Place it above Referral and use the group in acquisition and key event reports.

What regex tracks ChatGPT, Perplexity, Claude and Gemini in GA4?

Use chatgpt\.com|chat\.openai\.com|perplexity\.ai|www\.perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|copilot\.com|claude\.ai|chat\.deepseek\.com|grok\.com with the matches regex operator. Apply it to Source in a channel rule or to Session source in an Exploration filter. It also covers Copilot, Deepseek and Grok, and GA4 requires the full source value to match one alternative.

Why does ChatGPT traffic sometimes appear as Direct?

ChatGPT traffic lands in Direct when a visit arrives with no referrer and no utm_source tag. Without that evidence, neither the AI Assistant channel nor a custom rule identifies ChatGPT. Reclassifying data doesn’t recover a source that was never recorded.

Does Google count AI Overviews clicks as organic traffic?

Yes. Google’s Organic Search definition includes AI Overviews and AI Mode, and the AI Assistant channel excludes them. An assistant-domain regex won’t separate those clicks, so state that distinction in every AI report.

Does GA4's built-in AI Assistant channel track Perplexity and Claude retroactively?

Google’s release note adds Claude, but neither the channel definition nor that release note lists Perplexity, and the channel assigns visits from May 13, 2026. A custom channel group with a Source regex covers both assistants. Because custom groups apply retroactively, it also sorts earlier sessions that GA4 recorded with a source.

How do I measure conversions and key events from AI search traffic in GA4?

Mark the actions that matter as key events, which GA4 used to call conversions. View key events by your AI channel in Traffic acquisition and by First user source in User acquisition, then filter an Exploration by the same regex for landing-page detail. Keep key event definitions and date ranges identical across channels.