Most GA4 setups misattribute AI-referred sessions as “Direct” traffic. To fix this, create a custom AI channel group in GA4 Admin using a regex filter that captures ChatGPT, Perplexity, Gemini, Claude and Copilot as distinct sources — then build an Exploration report to track volume, engagement and conversions over time.
AI-referred web sessions grew 527% between January and May 2025. If your GA4 dashboard isn’t showing a corresponding rise in a clearly labelled AI traffic segment, you’re not measuring the fastest-growing acquisition channel in digital marketing — you’re just watching it disappear into “Direct.”
This guide walks you through the exact steps to set up AI traffic tracking in Google Analytics 4, what metrics actually matter, and how to connect those numbers to your GEO strategy.
Why AI traffic hides in GA4 by default
When a user clicks a link inside a ChatGPT response, the referral header is often stripped or passed as a dark social referral — meaning GA4 records it as Direct / (none) rather than attributing it to ChatGPT. The same is true for Perplexity in some browser and app contexts, and for Claude on mobile.
The result: your AI-driven awareness and interest is inflating your “Direct” numbers while your actual paid, organic and referral channels look artificially weaker by comparison. Your cost-per-acquisition calculations are wrong. You may be defunding the channels actually building awareness.
Step 1 — Create a custom AI channel group in GA4
A custom channel group lets you segment AI traffic natively inside every GA4 standard report without needing Explorations every time.
i. Open Channel Groups
Navigate to GA4 Admin → Data Display → Channel Groups. Click Create new channel group. Do not edit the Default Channel Grouping — that will alter your historical data permanently.
ii. Name the group
Use a clear label such as AI Search & LLMs or Generative AI Referrals. This name will appear in every report that uses channel groupings.
iii. Add a new channel called "AI Chatbots"
Click Add new channel. Set the dimension to Session Source, the operator to matches regex, then paste the regex string below.
iv. Paste the regex
This captures all major AI platforms including ChatGPT, Perplexity, Gemini, Claude and Copilot:
.*chatgpt\.com.*|.*chat\.openai\.com.*|.*perplexity\.ai.*|.*gemini\.google\.com.*|
.*claude\.ai.*|.*copilot\.microsoft\.com.*|.*bing\.com.*|.*you\.com.*|.*phind\.com.*
v. Prioritise the channel
Drag your new AI Chatbots channel to the top of the channel list so it takes priority over “Direct” when sources overlap.
vi. Save and allow 24–48 hours
Channel group changes only apply to data going forward. Set a recurring monthly reminder to review the source list and add new AI platforms as they emerge.
Step 2 — Build an AI Traffic Exploration report
For deeper analysis — including landing pages, conversion paths and new-vs-returning breakdowns — build a custom Exploration in GA4.
i. Open Explore → Blank
Name the exploration “AI Referral Traffic” and set a rolling 90-day date range.
ii. Add dimensions
Add: Session source / medium · Page referrer · Landing page + query string · New / returning.
iii. Add metrics
Add: Sessions · Engaged sessions · Engagement rate · Session duration · Conversions · Revenue (if applicable).
iv. Apply a segment filter
Create a segment where Session Source matches the same regex as your channel group. This isolates AI sessions from everything else.
The metrics that actually matter for GEO reporting
Not all GA4 metrics are equally useful when evaluating AI referral performance. Here’s what to focus on and why:
| Metric | Why it matters for GEO | What good looks like |
|---|---|---|
| Engaged sessions | AI visitors tend to arrive with high intent — they’ve already been “pre-qualified” by the AI | Engagement rate >65% (AI avg is typically higher than organic) |
| Landing page | Shows which content AI engines are citing and sending traffic to | Pillar articles and FAQs should dominate |
| Session source | Tells you which AI platforms are citing you most | Perplexity and ChatGPT typically lead; watch Gemini share grow |
| Conversion rate | The key commercial metric — AI traffic converting at ~14.2% vs ~2.8% for organic is the GEO business case | AI conversions should outperform any other acquisition channel per session |
| New users % | AI search is a top-of-funnel discovery channel — high new user % is expected and healthy | 70–85% new users is normal for AI referral traffic |
June 12, 2026
If AI traffic is still showing up as ‘Direct’ in your GA4, you’re not measuring GEO — you’re flying blind on your fastest-growing channel.
Tracking AI Referral Traffic in G4A
Step 3 — Set up conversion tracking for AI-referred sessions
The full value of GEO only becomes visible when you can link AI referrals to actual business outcomes. In GA4, this means:
- Marking your key goal events (form submissions, demo requests, purchases) as Conversions under Admin → Events.
- Adding a secondary dimension of Session source / medium to your conversion reports to see which AI engine drove each conversion.
- If you use Google Ads, importing these AI-attributed conversions to see how GEO content is supporting paid performance.
Step 4 — Build a monthly GEO reporting dashboard
Once your tracking is in place, standardise a monthly GEO report that your team and stakeholders can read in under five minutes. The core sections should be:
- AI traffic volume — total sessions from AI sources, MoM and YoY trend
- Top cited pages — which articles and pages AI engines are sending traffic to
- Top citing engines — ChatGPT vs Perplexity vs Gemini vs Claude share breakdown
- AI conversion rate — compared to organic, direct and paid benchmarks
- AI brand mentions — manual spot-check: ask each AI engine your top-3 category queries and record what it says
The last item — manual AI spot-checks — is not something GA4 can automate. It’s a qualitative gut-check that complements your quantitative data and is one of the most valuable 15 minutes you can spend each month on GEO health.
Glossary
- Dark social referral
- Traffic from links shared in closed or semi-closed environments (AI chat interfaces, messaging apps) where the referral header is stripped, causing GA4 to record it as Direct.
- Custom channel group
- A GA4 feature that lets you define custom traffic segments (e.g., “AI Chatbots”) that apply consistently across all standard reports.
- Engaged session
- A GA4 session that lasted 10+ seconds, had a conversion event, or had 2+ page views — replacing the old “bounce rate” metric as a quality indicator.
Read more in our comprehensive GEO Glossary
Frequently Asked Questions
Will the custom channel group affect my historical data?
No. Custom channel groups only apply to data collected after the group is created. Your historical “Default Channel Grouping” data is preserved separately.
Why does Perplexity sometimes not show up as a referral?
Perplexity uses both direct citations (which pass referral headers) and inline summaries (which may not). The regex filter captures clicks that do pass a header; dark-social Perplexity traffic may still appear as Direct. This is a platform-level limitation, not a GA4 configuration issue.
Which AI platforms should I add to my regex in 2026?
Core platforms to track: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com. Emerging platforms worth adding: you.com, phind.com, poe.com, and any new AI search product launched by your key industry platforms.
Is a 14.2% AI conversion rate realistic for my business?
The 14.2% figure comes from aggregated multi-site studies and varies significantly by industry and conversion definition. It is best used as a benchmark direction (AI traffic converts substantially better than organic) rather than a precise target. Set your own baseline once you have 90 days of clean AI referral data in GA4.