GA4 AI Assistant Channel: Measuring ChatGPT, Gemini, and Claude Traffic

GA4 now has a dedicated channel for AI assistant traffic. On May 13, 2026, Google Analytics added "AI Assistant" to the Default Channel Group, sessions from ChatGPT, Gemini, Claude, and other AI assistants are now classified there automatically, with no setup required. Open Traffic acquisition in an existing property and the AI Assistant row is already populated.^[5] The catch: the channel does not capture every AI-driven session. Three structural blind spots, referrer-stripped sessions, platforms absent from the list (Perplexity), and data before May 13, 2026, can distort any analysis that treats the channel number as the complete picture.

What the GA4 AI Assistant Channel Is

GA4 AI Assistant channel is a Default Channel Group entry in Google Analytics 4 that automatically classifies sessions arriving from recognized AI assistant referrer domains, chatgpt.com, gemini.google.com, claude.ai, and others, by setting medium = "ai-assistant" and campaign = "(ai-assistant)". It applies to all GA4 properties without any manual configuration.^[1]

Dark AI traffic refers to AI assistant sessions that reach a site without a referrer header. GA4 records these as (direct)/(none), so they never enter the AI Assistant channel. The primary causes are in-app link clicks, mobile app-to-browser transitions, in-app browsers (WebView), and private browsing.^[2]

UTM self-tagging is the practice of appending a utm_source= parameter to URLs published on AI-accessible surfaces (llms.txt, corporate blogs, etc.) to make AI-driven traffic identifiable even without a referrer header. ChatGPT Search applies utm_source=chatgpt.com automatically to outbound links, when that parameter is present, GA4 attributes the session to ChatGPT regardless of whether a referrer header was sent.^[3]

What GA4 Shows: and What It Doesn't

The table below maps each major AI assistant to its GA4 default channel classification.^[5][6]

How the Channel Works

When a visitor clicks through from a recognized AI assistant, the browser sends an HTTP Referer header that GA4 intercepts and inspects. If the referring domain appears on Google's AI assistant list, three session-level values are overwritten at once: Session medium becomes "ai-assistant", Session default channel group becomes "AI Assistant", and Session campaign becomes "(ai-assistant)".^[1][5]

The practical effect is that traffic from ChatGPT, Gemini, and other recognized platforms, previously scattered across Referral or swallowed into Direct, collapses into a single, named row. Any abrupt-looking shift in channel rows on May 13, 2026 reflects the new classification rule taking effect, not a retroactive change to historical data. The rollout was gradual; most properties were updated by early June 2026.

Previously, isolating AI traffic required hand-building a custom channel group with regex patterns for each domain. The new channel automates that, though platforms absent from Google's list still require manual configuration.^[4]

The Dark Traffic Problem

Referrer-stripped sessions are the channel's principal blind spot. ChatGPT's built-in Atlas browser strips the referrer header on outbound navigation; the same loss happens during mobile app-to-browser handoffs, in-app WebView contexts, and private browsing. GA4 assigns all of these to (direct)/(none) and counts them in the Direct channel instead.^[3]

Multiple published analyses document this gap as structurally significant, enough that treating the AI Assistant channel figure as a complete account of AI-driven visits systematically understates reality.^[2][4] ChatGPT Search does append utm_source=chatgpt.com to outbound links, which partially compensates, but only for web sessions, app-based traffic is unaffected.

The channel is also not retroactive. Sessions before May 13, 2026 remain in their original channels, Referral or Direct. To approximate AI traffic from that earlier period, use BigQuery Export and filter session_source for chatgpt.com, gemini.google.com, and similar domains.^[4]

Confirming the AI Assistant Channel in GA4

The channel requires no setup, here is how to read it.

Step 1, Open the report

In the GA4 property, go to Reports → Acquisition → Traffic acquisition.

Step 2, Switch the dimension

At the top of the table, set the dimension to "Session default channel group". This may already be the default.

Step 3, Find the AI Assistant row

Look for the "AI Assistant" row. Numbers populate from May 13, 2026 onward. If the row is absent, check whether the date range starts before that date.

Step 4, Break down by platform

Add "Session source" as a secondary dimension. This separates chatgpt.com, gemini.google.com, and other domains so you can see which AI assistant sends the most traffic.^[1]

Step 5, Map to conversion events

Compare conversion rates (purchases, form submissions, key events) for the AI Assistant channel against Organic Search and Direct. This is the metric that shows how AI-referred visits convert relative to other intent signals.

Adding Perplexity and Other Missing Platforms

Perplexity lands in Referral, not AI Assistant, because Google has not included it in the default channel definition.^[6] To pull it into your AI traffic totals, build a custom channel group: Admin → Data display → Channel groups → create a new channel, with the condition Session source contains "perplexity.ai". If claude.ai has drifted out of the default channel in your property, add it as a parallel condition in the same group.

The Perplexity optimization guide covers how Perplexity selects content for citation and how that affects traffic patterns.

Estimating Dark Traffic

Three approaches used together give a more complete picture of actual AI-driven visits than the channel number alone.^[2][4]

Approach 1, High-intent new-user Direct filter

In Explore → Free form, build a segment combining: Session default channel group = Direct, New/returning = New, and a conversion event condition (purchase, form submit, or key event). New users arriving directly on product or service detail pages are strong candidates for AI assistant referrals that lost their referrer header. Add that filtered session count to the AI Assistant channel total for an upper-bound AI traffic estimate.

Approach 2, Brand search cross-reference

Compare the AI Assistant channel session count week-over-week against branded query volume in Google Search Console. When both rise in sync, the implication is that total AI-driven visits, including the dark traffic portion, are larger than the GA4 channel shows. The GEO KPI measurement guide covers this cross-analysis method in detail.

Approach 3, CDN and server log comparison

Aggregate requests in your Nginx or CloudFront access logs where the Referer header contains chatgpt.com, claude.ai, or other AI domains. A server log count significantly higher than the GA4 AI Assistant number indicates a higher dark traffic ratio. Server logs only capture sessions that sent a Referer header, so they cannot quantify dark traffic directly, but they set a baseline for validating whether GA4's channel figures are in a plausible range.

Connecting AI Channel Data to GEO Strategy

The AI Assistant channel is not just a traffic metric. It is a citation signal, a record of which AI assistants are sending visitors from links they cited. Strong ChatGPT traffic suggests your content structure is working for ChatGPT retrieval. No Gemini traffic may indicate a gap in Google AI Overview optimization. Reading this alongside the multi-engine measurement guide supports channel-specific response strategies.

Where GA4's AI Assistant channel measures visitors who clicked a link inside an AI assistant, a dedicated solution like Brand Visibility Index (BVI) tracks upstream AI-surface exposure that never produces a click, the appearance, context, and rank of a brand's name inside AI responses before any visit occurs. BOIDA ↗ (operated by Designovel) covers major AI engines including ChatGPT, Claude, Gemini, and Perplexity, with Korean-language and domestic engine support, as an end-to-end platform spanning measurement, diagnosis, and execution. It launched in beta in December 2025 (as claimed) and holds an ACM CHI 2026 paper presentation and NVIDIA Inception membership. Tracking the three signals together, GA4 AI channel sessions, Search Console branded query trends, and AI visibility index (BVI), gives a complete view of the AI search conversion funnel.

See also: AI referral traffic landscape 2026 and IAB AI visibility 4P measurement standard.