LearnHow to Track AI Assistant Traffic in GA4
Measurement

How to Track AI Assistant Traffic in GA4

GA4 now has a native AI Assistant channel covering ChatGPT, Gemini, Copilot, DeepSeek, and Grok—but Perplexity and Claude need a custom channel group, and 35-70% of AI referrals arrive as Direct. Here's the step-by-step setup, and how to read the numbers honestly.

Jordan Hong Tai
Jordan Hong Tai
8 min readUpdated Jul 30, 2026
How to Track AI Assistant Traffic in GA4
TL;DR

GA4 now groups referrals from ChatGPT, Gemini, Copilot, DeepSeek, and Grok into a native "AI Assistant" channel automatically — but Perplexity and Claude are not included, and you need a custom channel group with a simple regex to capture them. More importantly: 35–70% of AI-originated visits arrive with no referrer and land in Direct, so whatever number GA4 shows you is a floor, not the truth. Set up the channel, read it as a minimum, and watch the trend.

Key Takeaways

  • GA4's native AI Assistant channel covers ChatGPT, Gemini, Copilot, DeepSeek, and Grok out of the box
  • Perplexity and Claude referrals are NOT in the default channel — you need a custom channel group with a regex on source
  • 35-70% of AI referrals arrive as Direct traffic, so treat your AI channel numbers as a floor
  • Setup takes about 10 minutes: Admin → Data settings → Channel groups → create an 'AI Assistants' channel

Why AI Assistant Traffic Deserves Its Own Channel

A growing share of buyers now start research inside an AI assistant instead of a search engine. When ChatGPT or Perplexity cites your site and the user clicks through, that visit is fundamentally different from organic search: the user arrived pre-qualified by a synthesized answer that already mentioned you, often after comparing you to competitors.

If those visits are lumped into Referral or Direct, you can't see whether your AI visibility work is producing actual sessions. Separating them into their own channel is the single highest-leverage analytics change for any team doing AEO.

What GA4 Tracks Natively

Google added an AI Assistant channel to GA4's default channel groups. It automatically classifies referral traffic from:

  • ChatGPT (chatgpt.com referrals)
  • Gemini (gemini.google.com)
  • Copilot (copilot.microsoft.com)
  • DeepSeek (chat.deepseek.com)
  • Grok (grok.com)

If your property is on a recent GA4 configuration, you'll see this channel appear in your Traffic acquisition reports with no setup at all. Check Reports → Acquisition → Traffic acquisition and search for "AI" in the channel column.

Two of the most important assistants for B2B research — Perplexity and Claude — are not in the native channel. Perplexity cites sources on nearly every answer, so for many brands it drives more click-throughs per mention than any other engine. If you skip the custom channel setup below, you are undercounting your best-converting AI source.

Step-by-Step: Add Perplexity and Claude via a Custom Channel Group

GA4 supports custom channel groups, which let you define your own classification rules on top of the default. Here's the full setup:

  1. Open Admin. In GA4, click the gear icon (bottom left).
  2. Go to Data settings → Channel groups. You'll see the Default Channel Group listed.
  3. Create a new channel group. Click "Create new channel group." GA4 copies the default rules as your starting point — don't start from scratch.
  4. Add a channel called "AI Assistants." Click "Add new channel" and name it. Naming matters: this label is what you'll see in every report.
  5. Set the condition. Match Source against a regular expression:
    perplexity|claude\.ai
    This catches perplexity.ai and www.perplexity.ai as well as claude.ai referrals. If you want one unified AI channel instead of GA4's split, extend the regex: chatgpt|perplexity|claude\.ai|gemini|copilot|deepseek|grok.
  6. Order the channel above Referral. Channel rules evaluate top-down; if Referral matches first, your AI channel never fires. Drag "AI Assistants" above Referral in the rule order.
  7. Save. Custom channel groups apply to data going forward and are available as a dimension in Traffic acquisition and Explorations.

10-minute checklist

  • ✓ Admin → Data settings → Channel groups
  • ✓ Create custom group (copies default rules)
  • ✓ Add channel "AI Assistants" with source matching regex perplexity|claude\.ai
  • ✓ Move it above Referral in rule order
  • ✓ Verify after a few days in Traffic acquisition with the custom group selected

The Direct-Traffic Problem: Your Numbers Are a Floor

Here's the honest caveat that most "track AI traffic" guides skip: a large share of AI-originated visits never carry a referrer at all. Native mobile apps, some in-app browsers, copied links, and certain privacy configurations strip the referrer — and those sessions land in Direct.

Across published analyses and our own client data, somewhere between 35% and 70% of AI-assistant-originated visits arrive as Direct. The range is wide because it depends on your audience's device mix and which assistants they use.

Practical implications:

  • Treat the AI Assistant channel as a minimum bound. If GA4 shows 200 AI sessions, the real number is plausibly 300–600.
  • Watch the trend, not the absolute. The undercount ratio is roughly stable month to month, so growth in the visible slice reflects real growth.
  • Correlate with unexplained Direct growth. If Direct traffic to deep content pages (not your homepage) is climbing alongside AI referrals, that's usually the invisible AI slice showing up.

How to Read the Numbers Once They Flow

In Reports → Acquisition → Traffic acquisition, switch the primary dimension to your custom channel group. Useful cuts:

  • Landing page by AI channel: which pages AI assistants actually send people to. These are the pages the engines find citation-worthy — study their structure.
  • Engagement rate and conversions by channel: AI-referred visitors typically arrive later in their research and engage more deeply than generic organic visitors.
  • Session source within the AI channel: which assistant drives what. Perplexity usually over-indexes on clicks; ChatGPT on volume.

Tracking Is the Floor. Growing It Is the Work.

GA4 tells you how many people arrived from AI assistants. It cannot tell you how often those assistants mention you, who they recommend instead, or which sources they cite when they answer questions in your category. That's the upstream funnel — and it's where the growth actually happens. AI brand monitoring covers the methodology; the short version is that you need a recurring, repeated, multi-model audit of your category questions.

Grow the traffic you just started tracking

CiteScore's AI Visibility Audit runs 100 prompts across ChatGPT, Claude, Gemini, and Perplexity, shows which sources AI cites for your category, and gives you a prioritized fix list. Weekly tracking with confidence-banded scores on every paid plan.

Frequently Asked Questions

Does GA4 track ChatGPT traffic automatically?

Yes. GA4's native AI Assistant channel groups referrals from ChatGPT, Gemini, Copilot, DeepSeek, and Grok automatically. Perplexity and Claude are not included by default — you need a custom channel group with a regex match on the source to capture them.

Why is my AI referral traffic so low in GA4?

Because a large share of it is invisible. Between 35% and 70% of visits that originate in an AI assistant arrive with no referrer and land in Direct. Treat the AI Assistant channel number as a floor, not the true total, and watch the trend rather than the absolute.

How do I track Perplexity and Claude traffic in GA4?

Create a custom channel group (Admin → Data settings → Channel groups) with a channel called "AI Assistants" whose condition matches source against the regex perplexity|claude\.ai, ordered above the Referral rule. Combined with GA4's native AI channel, this covers all the major assistants.

References & Further Reading

  1. [1]GA4 default channel groupsGoogle
  2. [2]GA4 custom channel groupsGoogle
  3. [3]GA4 traffic-source dimensionsGoogle

About the author

Jordan Hong Tai

Jordan Hong Tai

LinkedIn

CEO & Founder, CiteScore

Jordan Hong Tai is the founder of CiteScore. He works with brands on how AI assistants like ChatGPT, Perplexity, Gemini, and Claude discover, cite, and recommend them.

Related Articles

Check your AI visibility

Run a Brand AEO Audit to see where your brand appears in AI answers—and get a content plan to improve it.