How to track AI chatbot traffic in GA4: 3 methods (2026)

GA4 added a native AI Assistant channel in May 2026. That is useful. It is also incomplete by design: it only fires when the referrer header is intact, and a lot of AI sessions never send one. The ChatGPT desktop app, most mobile apps, and any platform not on Google’s recognized list all land somewhere else, usually in direct traffic or generic Referral.

This post covers how to track AI chatbot traffic across all three buckets, ranked by setup time.

Why GA4 fails to track AI chatbot traffic by default#

When someone clicks a link inside an AI assistant, the browser sends a referrer header. GA4 reads that header and attributes the session. When the header is missing, GA4 defaults to direct traffic.

Two things strip referrer headers consistently:

• Native apps (ChatGPT desktop, Claude desktop, mobile apps on both platforms)
• Certain browser privacy configurations

Beyond stripped headers, the native AI Assistant channel only recognizes sources Google has explicitly added. As of June 2026, Perplexity is not confirmed on that list. Perplexity referrals with intact headers still land in generic Referral.

The result: your AI chatbot traffic is split across three buckets in GA4, and that split is invisible unless you look for it.

Method 1: GA4’s native AI Assistant channel (0 min setup)#

On May 13, 2026, Google confirmed in its Analytics release notes that traffic from recognized AI assistants is now auto-tagged with medium = ai-assistant and grouped under a dedicated "AI Assistant" channel in the Default Channel Group. No configuration required across any GA4 property.

Where to find it: Reports > Acquisition > Traffic acquisition. Make sure the channel grouping dropdown shows "Default channel group." Look for the AI Assistant row.

Three things to know before you rely on this number:

1. It only fires when the referrer header is intact.
2. Google has confirmed ChatGPT, Gemini, and Claude as recognized platforms. Perplexity’s inclusion is unconfirmed as of this writing. Check your own reports to see whether Perplexity sessions appear under AI Assistant or still show as Referral.
3. It is not retroactive. Historical data before May 13, 2026 stays wherever GA4 originally classified it. Broad rollout across properties completed around June 7, 2026, so properties that received the update late may show limited data.

Use this when you want a quick trend read on ChatGPT, Gemini, and Claude from May 2026 onward. It is not sufficient for pre-May data or for full platform coverage.

Method 2: track AI chatbot traffic in Explore with regex (10 min setup)#

This is an updated version of the original approach from this post. Even with the native GA4 AI Assistant channel live, you need this method for:

• Any data before May 13, 2026
• Platforms not confirmed in the native channel (Perplexity, DeepSeek, Grok, Meta AI)
• Side-by-side platform comparison in a single view

Setup:

1. Go to Explore and create a new blank exploration.
2. Add dimensions: Session source and Landing page.
3. Add metrics: Sessions and Engaged sessions.
4. Drag Session source to the Rows section.
5. In the Filters section, choose Session source > matches regex, then paste the pattern below.
6. Name it "AI Traffic Sources" and save.

One change from the original version of this article: I switched from "Page referrer" to "Session source" as the primary dimension. Page referrer captures every referrer within a session, which can overcount. Session source gives you one attribution per session, matching how every other channel report works in GA4.

Updated 2026 regex:

1chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|deepseek\.com|grok\.com|meta\.ai|chat\.mistral\.ai

Review this list quarterly. New platforms launch regularly, and the ones not on your list are invisible.

Platform referral behavior in GA4:

• ChatGPT - chatgpt.com, chat.openai.com - confirmed in native channel - appends utm_source=chatgpt.com on some links
• Gemini - gemini.google.com, bard.google.com - confirmed - no UTM parameters
• Claude - claude.ai - confirmed - desktop app strips referrer
• Perplexity - perplexity.ai - not confirmed in native channel - browser sessions pass referrer reliably
• Microsoft Copilot - copilot.microsoft.com - unconfirmed - Bing-integrated sessions may differ
• DeepSeek - chat.deepseek.com - not in native channel - no UTM parameters
• Grok - grok.com - not in native channel - no UTM parameters
• Meta AI - meta.ai - not in native channel - no UTM parameters

From my own GA4 data on k-o.pro, ChatGPT, Perplexity, Gemini, and DeepSeek show up consistently. If you publish SEO or analytics content, expect a similar distribution.

If you want to automate the analysis rather than check it manually, I covered a Claude-based workflow for this in Claude for SEO: 3 workflows I've automated as a practitioner.

Method 3: custom channel group AI traffic (15 min setup)#

The native channel auto-classifies some AI traffic. The Explore method gives you historical data. A custom channel group for AI traffic does both: it permanently surfaces all AI sessions across your standard reports, not just in Explore, and it covers platforms the native channel misses.

Setup:

1. Open Admin (gear icon) > Data display > Channel groups.
2. Click Create new channel group. Do not edit the default one.
3. Name the group "AI Traffic (All Sources)".
4. Click Add new channel, name it "AI Search".
5. Set the rule: Session source > matches regex, then paste the 2026 regex above.
6. Drag the "AI Search" channel above "Referral" in the priority list.
7. Click Save.

Step 6 is the one most setups get wrong. GA4 evaluates channel rules top-down. If "Referral" sits higher in the list, AI sessions with valid referrers get classified as generic Referral before your rule runs.

Once live, select this channel group from the dropdown in your standard Traffic acquisition report. Expect 24-48 hours before data starts populating.

Limitation: this applies going forward only. For data before you created the group, you still need Method 2.

One addition worth making alongside this setup: for any links you control that AI tools might cite, add UTM parameters. Sessions with UTM tags are attributed correctly regardless of what the referrer header does. Use the UTM Campaign URL Builder to build consistent UTM strings.

Which method to use#

• Native AI Assistant channel - 0 min setup - covers ChatGPT, Gemini, Claude (referrer intact only) - no historical data, May 2026 onward - best for a quick trend check
• Explore + regex - 10 min setup - covers all platforms, any date range - yes historical data - best for retroactive analysis and full platform breakdown
• Custom channel group - 15 min setup - covers all platforms ongoing - no historical data - best for permanent reporting across all GA4 reports

Run Method 2 first to see what you already have. Then set up Method 3 to capture everything going forward without manual effort.

What to do with your AI chatbot traffic data#

Once AI sessions are visible, two things are worth doing immediately.

Add Landing page as a dimension in your Explore report. The pages showing up most often are the ones AI assistants are actively citing. Those pages tend to share structural traits: direct definitions, numbered steps, comparison tables. Check whether your other content has the same patterns.

Compare engagement for AI sessions versus your organic baseline. Adobe's data on US retail sites found AI-referred traffic converted 42% better than non-AI traffic in March 2026, with a 12% higher engagement rate. The general pattern holds across content sites: these visitors arrive with context the AI already gave them, which changes how they behave.

The most actionable gap is pages that rank in traditional search but get zero AI chatbot traffic. The cause is usually structural, not topical. Those pages lack extractable definitions, comparison tables, or direct answers. That is a content edit, not a technical fix.

Also worth checking while you are at it: whether AI crawlers can actually reach your content in the first place. The AI Crawler Access Checker lets you verify which AI bots your robots.txt allows.

FAQ#

Does GA4 automatically track AI traffic now?#

Since May 13, 2026, GA4 auto-classifies traffic from ChatGPT, Gemini, and Claude under a native "AI Assistant" channel with no setup required. But sessions without referrer headers (most desktop and mobile app direct traffic) still land in Direct. Perplexity's classification in the native channel is unconfirmed. For complete coverage you still need a custom channel group with regex filtering alongside the native channel.

Why does ChatGPT show as direct traffic in GA4?#

ChatGPT's desktop and mobile apps do not pass referrer headers when opening external links. GA4 sees no referrer and defaults to direct traffic. The browser-based version of ChatGPT generally passes the header correctly. You can partially compensate by filtering for sessions with utm_source=chatgpt.com, which ChatGPT appends to some outbound links.

What regex pattern should I use to track AI traffic in Google Analytics?#

Use this on Session source in Explore or as the matching rule in a custom channel group:

1chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|deepseek\.com|grok\.com|meta\.ai|chat\.mistral\.ai

Review it quarterly as new platforms launch.

Can I see which queries brought traffic from ChatGPT or Perplexity in GA4?#

No. GA4 records the referrer domain, not the prompt the user typed. You can see which landing pages receive AI chatbot traffic, which tells you which content is being cited. For prompt-level insight, query the AI tools directly with your target topics and observe whether your content appears.

Does the custom channel group work retroactively?#

No. Custom channel groups classify sessions from the creation date forward. For historical data, use the Explore method with a regex filter on Session source, which you can apply to any date range in your GA4 property.

Where to go from here#

The native GA4 AI Assistant channel makes the basic picture visible with zero work. For the full picture, you need the Explore report to cover historical data and the platforms not on Google's list, and a custom channel group to keep everything surfaced permanently.

Once the data is readable, the real work starts: which pages get cited, whether AI sessions convert differently from organic, and which content is invisible to AI despite ranking in classic search. That last gap is usually a structural content problem with a concrete fix.

If your next question is what to do after you have identified the gap, also check whether AI crawlers can access your content at the AI Crawler Access Checker and look at making your pages more machine-readable from the ground up.