The Agency Guide to Tracking AI Traffic in GA4 — Setup, Regex Patterns, and More

Your clients are getting traffic from ChatGPT, Perplexity, and Claude right now. The problem? Google Analytics 4 is hiding it from you.

Here’s what’s happening: someone asks ChatGPT for product recommendations in your client’s industry. The AI responds with a helpful answer and includes a link to your client’s website. That visitor clicks through, browses three pages, and converts. In your GA4 reports, this high-value session appears as just another “Referral” visit—lumped together with traffic from random forums, coupon sites, and everywhere else.

This matters more than you might think. According to Adobe Analytics research from late 2025, AI referral traffic converts at 4.4 times the rate of traditional search traffic. These visitors spend 68% longer on sites and view over three times as many pages per session. Yet without custom configuration, most agencies can’t tell their clients how much AI-driven traffic they’re actually getting—or which AI platforms are sending it.

This guide will show you exactly how to uncover that hidden traffic. We’ll walk through the GA4 setup, share the regex patterns that actually work in 2025-2026, and explain why even a perfect configuration will still miss a significant portion of AI-influenced visits. By the end, you’ll have everything you need to track AI traffic accurately—and just as importantly, you’ll understand the limitations so you can set proper expectations with clients.

Understanding the Difference Between AI Crawlers and AI Referrals

Before touching any settings in GA4, you need to understand a distinction that trips up many marketers: AI traffic comes in two completely different forms, and they require different tracking approaches.

AI Crawlers: The Bots You Don’t Need to Track in GA4

The first type is crawler traffic. These are automated bots like GPTBot, ClaudeBot, and PerplexityBot that visit your site to scrape content. They’re gathering training data for large language models or indexing pages for AI-powered search features. No human is involved in these visits.

GA4 automatically filters out most of this bot traffic using the IAB International Spiders and Bots List. This happens behind the scenes—you can’t turn it off, and you can’t see how much traffic was filtered. That’s actually fine for marketing purposes, because crawler visits have zero engagement metrics that matter. A bot doesn’t read your content, click your calls-to-action, or convert into a lead.

If you need visibility into crawler activity for SEO or content strategy purposes (which AI bots are accessing your site, and how often?), you’ll need to analyze your server logs using tools like Screaming Frog. That’s a separate discipline from what we’re covering here.

AI Referrals: The Human Traffic You Must Track

The second type—and the one that should command your attention—is referral traffic. This is what happens when a real person uses an AI assistant, receives a response that includes a link to your client’s website, and clicks through to visit. These are genuine prospects engaging with genuine content, and they exhibit remarkably strong engagement signals.

According to the November 2025 Conductor benchmark report, 73% of visitors from AI referrals convert during their first session, compared to just 23% of visitors from Google organic search. The quality gap is stark—and closing fast. Revenue per visit from AI traffic improved from just 3% of non-AI traffic in July 2024 to 70% by May 2025, based on Adobe’s tracking across retail sites.

The problem is that GA4 doesn’t distinguish these high-value visitors. When someone clicks a link from ChatGPT’s response, GA4 sees “chatgpt.com” as the referrer and files that session under the generic “Referral” channel. It’s technically accurate, but practically useless—you can’t optimize a channel you can’t see.

Why GA4’s Default Setup Fails to Capture AI Traffic

Google Analytics 4 ships with 18 default channel groupings: Organic Search, Paid Search, Direct, Referral, Social, Email, and so on. Notice what’s missing? There’s no “AI” channel.

This creates three specific problems for agencies trying to measure AI-driven traffic:

All AI referrals get lumped into “Referral.” Traffic from ChatGPT sits alongside traffic from Yelp, Wikipedia, and obscure backlinks. You’d need to manually filter by source in every report to see AI traffic specifically. Most agencies never do this, so the data sits there unseen.

Google AI Overviews traffic is invisible. When someone clicks through from Google’s AI-generated answers at the top of search results, GA4 classifies this as regular “Organic Search.” There’s no automatic way to separate AI-assisted search clicks from traditional organic clicks. (We’ll cover a partial workaround later, but it has limitations.)

A significant portion appears as “Direct.” Mobile app traffic from ChatGPT often loses its referrer header entirely. When someone uses the ChatGPT iOS or Android app and clicks a link, the browser that opens may not know where the click originated. The result? GA4 logs it as Direct traffic, which is the analytics equivalent of “we have no idea where this came from.”

Google has acknowledged the need for AI traffic tracking. In late 2024, they updated their custom channel groups documentation to include an example regex pattern for “AI assistants.” But it remains an example buried in help documentation, not a default configuration. If you want your clients’ reports to show AI traffic, you need to set it up yourself.

Setting Up a Custom Channel Group for AI Traffic

The solution is to create a custom channel group that identifies AI referrers before they get classified as generic referral traffic. This isn’t complicated, but it does require careful attention to a few details that are easy to overlook.

Step-by-Step Configuration

Start by navigating to Admin → Data Display → Channel Groups in your GA4 property. You’ll see your existing Default Channel Group here.

Click “Create new channel group” to build a custom version. Give it a clear name like “Default + AI Traffic” so you can easily identify it in reports. Alternatively, you can duplicate the default group and modify the copy—either approach works.

Within your new channel group, click “Add new channel” and name it “AI Traffic” or “AI Referrals.” This is what will appear in your reports.

Now configure the matching condition. Set the field to Source, the operator to matches regex, and paste in your regex pattern (we’ll provide recommended patterns in the next section).

Here’s where most agencies make their first mistake: forgetting to reorder the channels. GA4 processes channel rules from top to bottom, assigning traffic to the first matching channel it finds. If your AI Traffic channel sits below the Referral channel, a visit from chatgpt.com will match Referral first and never reach your AI Traffic rule.

Click the “Reorder” button and drag your AI Traffic channel above the Referral channel. This ensures AI sources get captured by your custom channel before the generic Referral catch-all.

Save your channel group, then remember to apply it to your reports and explorations. Custom channel groups don’t automatically replace the default in existing reports—you need to select them specifically.

The Regex Patterns That Actually Work

The regex pattern you use determines which AI platforms your channel will capture. Google’s example pattern from their documentation covers the major players, but the community has developed more comprehensive alternatives.

Production-ready pattern (covering 30+ AI platforms as of January 2026):

^(?:chatgpt\.com|chat-gpt\.org|claude\.ai|quillbot\.com|openai\.com|blackbox\.ai|perplexity(?:\.ai)?|copy\.ai|jasper\.ai|copilot\.microsoft\.com|gemini\.google\.com|(?:\w+\.)?mistral\.ai|deepseek\.com|edgepilot|edgeservices|nimble\.ai|iask\.ai|aitastic\.app|bnngpt\.com|writesonic\.com|exa\.ai|cohere\.ai|huggingface\.co|anthropic\.com|poe\.com|pi\.ai|phind\.com|character\.ai|you\.com|meta\.ai|grok\.x\.com|x\.ai)$

This pattern was developed and is maintained by Dana DiTomaso at Analytics Playbook. It’s updated regularly as new AI platforms emerge and referral patterns change.

For agencies that want more granular reporting, consider splitting this into two channels: one specifically for ChatGPT (which accounts for roughly 87% of all AI referral traffic, according to Conductor’s research) and another for all other AI platforms. This approach lets you report on ChatGPT’s impact specifically while still capturing the long tail of smaller AI tools.

ChatGPT-only pattern:

^(?:chatgpt\.com|chat-gpt\.org|openai\.com|chat\.openai\.com)$

Other AI platforms pattern:

^(?:claude\.ai|anthropic\.com|perplexity(?:\.ai)?|gemini\.google\.com|copilot\.microsoft\.com|meta\.ai|poe\.com|you\.com|phind\.com|mistral\.ai|deepseek\.com|grok\.x\.com)$

A note on case sensitivity: GA4’s regex matching is case-sensitive by default. The sources should appear in lowercase in your data, but you can add the (?i) flag at the beginning of your pattern to make it case-insensitive if you’re seeing mixed case issues.

Complete AI Referrer Reference List

The AI landscape evolves quickly. Platforms launch, rebrand, and change their referral behavior without notice. Here’s a comprehensive list of current AI referrers as of January 2026, along with what you should know about each:

This list will need periodic updates. New AI tools emerge regularly, and existing platforms sometimes change their domain structure. Building a quarterly review of your regex patterns into your analytics maintenance routine is good practice.

The “Dark AI Traffic” Problem (And Why Your Numbers Will Always Be Incomplete)

Here’s an uncomfortable truth: even with perfect GA4 configuration, you’re only seeing a fraction of AI-influenced traffic. Industry estimates suggest visible AI referrals represent just 30-40% of actual AI-driven visits.

The rest disappears into what some analysts call “dark AI traffic”—visits influenced by AI that analytics tools simply cannot attribute correctly. Understanding why this happens will help you set realistic expectations with clients.

Copy-and-Paste Behavior

Most ChatGPT users don’t click links directly. They copy the URL from the response and paste it into their browser. This creates a new browser session with no referrer information—GA4 logs it as Direct traffic. The AI initiated the visit, but there’s no technical trail connecting them.

Mobile App Referral Loss

When someone uses ChatGPT’s mobile app and taps a link, the app opens the device’s default browser. This app-to-browser handoff often drops the referrer header entirely. The visit appears in GA4, but the source shows as Direct because the browser doesn’t know the click originated from ChatGPT.

AI Answers Without Links

Sometimes AI assistants mention brands or products without including hyperlinks. The user then types the company name directly into their browser (Direct) or searches for it on Google (Organic Search). The AI created the awareness and intent, but the resulting visit shows up under a completely different channel.

This is particularly common with informational queries. If someone asks “what’s the best CRM for small businesses” and ChatGPT mentions HubSpot by name without a link, that person might visit hubspot.com directly or Google “HubSpot pricing.” The AI-initiated journey becomes invisible to analytics.

Privacy Features Strip Referrers

Safari’s Intelligent Tracking Prevention and various browser privacy settings can strip or truncate referrer data. A visit that originated from an AI platform might arrive at your site with that origin information already removed.

What This Looks Like in Practice

The SaaS company Tally discovered this gap firsthand. They added a simple post-signup survey asking “How did you hear about us?” and found that 25% of new users mentioned ChatGPT—far exceeding what their analytics showed for AI referral traffic. The lesson: GA4 data should be considered a floor, not a ceiling, when estimating AI influence.

For agencies, this creates an opportunity to provide additional value. Implementing post-conversion surveys that specifically include AI options (“ChatGPT,” “Perplexity,” “An AI assistant,” etc.) gives you a second data point to triangulate against your GA4 numbers. When you can show clients that AI influence extends well beyond what appears in standard reports, you’re demonstrating analytical sophistication that competitors may lack.

Advanced Tracking: Google AI Overviews and Text Fragments

Google’s AI Overviews—the AI-generated answers that appear at the top of search results—present a unique tracking challenge. These clicks come from google.com, so GA4 classifies them as Organic Search. There’s no native way to separate AI-assisted search from traditional organic.

However, there’s a partial workaround using URL fragments. When someone clicks on highlighted text within an AI Overview (or Featured Snippet, or People Also Ask), Google appends a text fragment to the URL that looks like #:~:text=highlighted%20phrase. This fragment tells the browser to scroll to and highlight that specific text on the page.

You can capture this fragment using Google Tag Manager. Create a Custom JavaScript variable that checks for the presence of :~:text= in the URL hash:

function() {  var hash = window.location.hash;  if (hash.indexOf(':~:text=') !== -1) {    return decodeURIComponent(hash.split(':~:text=')[1]);  }  return null;}

When this variable returns a value, fire a custom event (something like ai_overview_click) and send the highlighted text as an event parameter. Register this parameter as a custom dimension in GA4, and you’ll be able to filter Organic Search traffic to see only sessions that arrived via text-highlighted links.

A few caveats: This technique captures highlighted-text clicks specifically, not all AI Overview clicks. Users who click links that don’t have text highlighting won’t trigger this detection. And the text fragment doesn’t tell you definitively that the click came from an AI Overview—Featured Snippets and PAA boxes use the same mechanism. Still, it provides useful directional data that most competitors won’t have.

GTM Configuration for Enhanced AI Tracking

Beyond the basic custom channel setup, Google Tag Manager allows you to capture AI referral data as event parameters. This gives you more flexibility for analysis and lets you create audiences based on AI traffic.

Create a Custom JavaScript variable that detects AI referrers:

function() {  var referrer = document.referrer.toLowerCase();  var aiPatterns = [    'chatgpt.com', 'chat.openai.com', 'openai.com',    'perplexity.ai', 'claude.ai', 'anthropic.com',    'gemini.google.com', 'copilot.microsoft.com',    'meta.ai', 'poe.com', 'you.com', 'phind.com',    'mistral.ai', 'deepseek.com', 'grok.x.com'  ];    for (var i = 0; i < aiPatterns.length; i++) {    if (referrer.indexOf(aiPatterns[i]) !== -1) {      return aiPatterns[i].split('.')[0];    }  }  return null;}

This variable returns the AI platform name (“chatgpt”, “perplexity”, “claude”, etc.) when detected, or null for non-AI traffic. You can then:

• Create a trigger that fires on Page View when the variable is not null
• Fire a custom event (like ai_referral_detected) with the platform name as a parameter
• Register ai_source_platform as a custom dimension in GA4
• Build audiences of “users who arrived via AI referral” for remarketing or analysis

This approach complements rather than replaces the custom channel group. The channel group handles attribution in standard reports; the custom event gives you flexibility for exploration and audience building.

The Ten Mistakes Agencies Make When Tracking AI Traffic

After working with dozens of analytics implementations, patterns emerge in what goes wrong. Avoid these common pitfalls:

1. Never creating a custom channel at all. This is the most common issue. Without explicit configuration, AI traffic remains hidden in the Referral bucket forever. GA4 won’t change its defaults for you.
2. Forgetting to reorder channels. The second most common mistake. Your AI Traffic channel must appear above Referral in the priority order, or traffic will never reach it.
3. Case-sensitivity issues in regex. GA4 regex is case-sensitive. If your pattern uses “ChatGPT” but the source appears as “chatgpt”, it won’t match. Stick to lowercase patterns.
4. Only tracking ChatGPT. While ChatGPT dominates current traffic, Perplexity, Claude, and especially Gemini (up 388% year-over-year) represent growing sources. A narrow pattern means missing emerging traffic.
5. Assuming GA4’s bot filter handles AI tracking. The bot filter removes crawler traffic—it doesn’t help identify valuable AI referral traffic. These are separate concerns.
6. Confusing crawlers with referrals. Bots scraping your site for training data and humans clicking links from AI responses require completely different analysis approaches. Don’t conflate them in reporting.
7. Using last-click attribution exclusively. AI often functions as an awareness touchpoint early in the customer journey. Last-click attribution may credit the conversion to a later touchpoint while AI did the heavy lifting. Data-driven attribution reveals a more accurate picture.
8. Ignoring Google AI Overviews. This traffic appears as regular Organic Search without special tracking. If you’re not implementing the text fragment detection described earlier, you have no visibility into AI-assisted search behavior.
9. Missing the “(not set)” medium problem. ChatGPT sometimes passes utm_source without utm_medium. This can cause traffic to be classified as “Unassigned” rather than matching your AI channel. Consider adding a condition that matches your AI source regex when medium equals “(not set)”.
10. Treating GA4 numbers as complete. Visible AI referrals are the tip of the iceberg. Set client expectations appropriately—the real AI influence is likely 2-3x what reports show.

Current Benchmarks

Setting appropriate expectations requires knowing what typical AI traffic looks like. Here are the current benchmarks from industry research:

Sources: Conductor November 2025 Report, Adobe Analytics Research (2025)

The key insight from these benchmarks: AI traffic is still small in absolute terms (1-3% of total) but dramatically outperforms on engagement and conversion metrics. This isn’t a volume play yet—it’s a quality play. Agencies that can demonstrate this quality difference to clients have a compelling story to tell about why AI traffic measurement matters.

Client Reporting Framework for AI Traffic

Once you’ve implemented tracking, you need to present the data in a way that’s meaningful to clients. Here’s a recommended structure for AI traffic reporting:

Overview Section

Start with total AI sessions from your custom channel, month-over-month and year-over-year trends, and a platform breakdown showing which AI tools are sending traffic. Include a comparison table showing AI traffic versus Organic Search versus Direct—this contextualizes the AI numbers within the broader traffic mix.

Engagement Quality Section

This is where AI traffic typically shines. Show session duration by source, pages per session, engagement rate, and bounce rate. Create a side-by-side comparison with other channels to highlight the quality difference. If AI visitors are spending twice as long on site and viewing three times as many pages, that’s a story worth telling.

Conversion Analysis Section

Report on key event rates (form submissions, signups, purchases) by AI platform. If you’re using data-driven attribution, show where AI appears in conversion paths—you may find it’s generating awareness that leads to conversions attributed to other channels. For e-commerce clients, include revenue attributed to AI traffic.

Content Performance Section

Identify which landing pages receive the most AI traffic. This reveals what content AI tools are citing when they send users to your client’s site. The pages that perform well organically may not be the same ones that AI tools prefer to reference—spotting these differences creates content optimization opportunities.

Looker Studio Formula for Custom Reports

If you’re building reports in Looker Studio and want to add an AI traffic dimension without creating a full custom channel group, use this calculated field:

CASE   WHEN REGEXP_MATCH(Session source, "chatgpt|openai|perplexity|claude|gemini|copilot|anthropic|deepseek|meta\\.ai")   THEN "AI Traffic"   ELSE Session default channel group END

This formula lets you segment traffic on the fly within Looker Studio reports without modifying the underlying GA4 configuration.

Implementation Checklist

Use this checklist to ensure complete AI traffic tracking implementation:

Phase 1: Quick Setup

These five steps get AI traffic visible in your GA4 reports. Most agencies can complete all of them in a single sitting—30 to 60 minutes total.

Result: AI referral traffic appears as its own channel in reports. You can see volume, engagement, and conversions from AI platforms immediately.

Phase 2: Advanced Tracking

Once the basics are in place, these steps give you deeper insight and fill gaps in attribution. Tackle them in any order based on what matters most for your clients.

Result: You have platform-level AI attribution, visibility into Google AI Overviews, a self-reported data layer to cross-reference against analytics, and client-ready dashboards.

Ongoing Maintenance

AI platforms change domains, launch new products, and alter referral behavior without notice. Build these habits into your regular analytics routine:

• Review and update regex patterns quarterly. New AI tools emerge regularly—check for platforms sending traffic that your pattern doesn’t yet capture.
• Set up automated alerts for significant changes in AI traffic volume (spikes or drops may signal new platform behavior).
• Cross-reference survey data with GA4 data to estimate the true scale of AI influence. GA4 numbers are the floor, not the ceiling.
• Monitor for case-sensitivity issues and (not set) medium problems that can cause AI traffic to slip into Unassigned.

What Actually Matters

Three numbers tell the story: AI referrals convert at 4.4× the rate of organic search, but visible AI traffic only accounts for 30–40% of actual AI-driven visits, and ChatGPT alone sends 87% of it.

Your GA4 setup needs exactly two things to capture this: a custom channel group with the regex pattern from this guide, and that channel ordered above Referral. That’s 30 minutes of work. Everything else — GTM events, AI Overview tracking, survey data — builds on that foundation.

The one thing GA4 can’t fix: most AI-influenced visits will never show up as AI traffic. Users copy-paste URLs, mobile apps drop referrers, and AI mentions brands without linking. A post-conversion survey asking “How did you hear about us?” with AI-specific options will consistently reveal 2–3× more AI influence than your analytics reports show.

Update your regex quarterly. AI platforms rebrand, launch new domains, and change referral behavior without warning. The pattern that works today will have gaps by next quarter.