The AI referrer hostname list, with a ready GA4 regex (2026)

By Nikhil Vemu · October 25, 2026 · Updated November 5, 2026

To group AI-assistant traffic in Google Analytics 4, match the referrer hostnames — chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and the rest — in a custom channel group. Below is the complete list as of October 2026, a copy-paste regex, and what GA4's own AI Assistant channel still files somewhere else.

These hostnames are the surfaces a person clicks a link from to reach your site. A large share of AI visits still arrive with no referrer at all — a copied link, or an in-app webview that strips the header — and those land in Direct on every tool. No regex recovers them. This list is for the part you can see.

The AI referrer hostnames, October 2026

This is the exact set Simplytics files into its AI channel. A GA4 property needs the regex below to do the same.

The GA4 regex to group them

Paste this into a custom channel group as a Source matches regex condition:

(^|\.)(chatgpt\.com|chat\.openai\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|edgeservices\.bing\.com|you\.com|poe\.com|phind\.com|deepseek\.com|chat\.deepseek\.com|grok\.com|meta\.ai|mistral\.ai|chat\.mistral\.ai|kimi\.com|qwen\.ai)$

Then build the group:

1. In GA4, open Admin → Data display → Channel groups and create a new group (it clones the default).
2. Add a channel named AI assistants.
3. Set its condition to Source matches regex and paste the string above.
4. Drag the AI assistants channel above Organic Search and Referral, so gemini.google.com and perplexity.ai match as AI first.
5. Save, then confirm in a report or DebugView.

A few things that trip people up. GA4's Source is the hostname, not the path, so bing.com/chat can never match — use the hosts above. GA4's "matches regex" is a partial match, so the (^|\.)…$ anchors are doing real work: they still catch www.perplexity.ai, but they stop a hostname like foryou.com from matching you.com. And phind.com shut down in January 2026 while bard.google.com now redirects to Gemini — both stay in the list for old links, so drop them if you want a shorter regex. The bare roots openai.com, deepseek.com, mistral.ai and qwen.ai can catch docs or API visits, not just chat; trim those too if your reports show noise.

What GA4's native AI Assistant channel still misses

Since 13 May 2026, GA4 has shipped a default AI Assistant channel, so some of this is now automatic. Google's documentation names ChatGPT, Gemini, DeepSeek, Copilot and Grok as examples; the full recognized list is unpublished and expanding. The custom group above still earns its place, because the native channel has fixed gaps:

Perplexity falling into Referral is the one most people miss, because it is the assistant that most reliably sends a referrer — the traffic is there to count, GA4 just files it elsewhere. AI Overviews and AI Mode arrive on google.com, so no referrer rule can split them from Organic Search. The full breakdown of what the native channel hides is in what GA4's AI Assistant channel still hides, and the step-by-step for the regex approach is in how to see ChatGPT, Perplexity and Gemini traffic in your analytics.

Which assistants actually send the referrals

Most of the list is small traffic. In SE Ranking's June 2026 study of 101,574 sites, AI was about 0.32% of all web traffic, and of the measurable AI referrals ChatGPT took roughly 75%, Gemini 12%, Perplexity 7%, Copilot 3.5% and Claude 2.6%. Contentsquare's 2026 Digital Experience Benchmark put AI-referred traffic up about 632% year on year, but still near 0.2% of visits. Both are third-party panels measuring different things — cite each with its source, and remember neither is your data. Your own referrer log is, which is the point of grouping these in the first place. Reading that mix well is its own trap: see is ChatGPT losing traffic share in 2026?.

The shortcut: the list ships in the tool

Simplytics is a cookie-less analytics tool that files ChatGPT, Perplexity, Claude, Gemini and the rest of this list into an AI channel automatically, matched before Search so gemini.google.com never leaks into Google. There is no regex to maintain and no channel group to build — it is a ~1.2 KB script, and it costs $1/month billed yearly ($12/year), against the $9–15/month of hosted privacy tools like Plausible and Fathom. You can see the Channels table on the live demo, and the full trade-off against GA4 is in the Google Analytics comparison.

When another tool fits better

• You run paid ads. GA4's Source Platform dimension and paid-vs-organic attribution do more than a referrer channel ever will.
• You already use GA4 and only want AI grouped. The regex above is the whole job — no new tool needed.
• You want the stripped-referrer clicks back. No tool can recover them; they land in Direct everywhere. Do not pay for a "recovery" feature that claims otherwise.

Last verified: 2026-10-25.