Blog/Article

How to Track AI Referral Traffic in GA4 (ChatGPT, Perplexity, Gemini) and Convert It

Why AI referrals from ChatGPT, Perplexity, and Gemini hide inside GA4 defaults, the exact custom channel group regex to surface them, and how to convert them.

JMJimit Mehta · 9 min read
Analytics dashboard concept representing AI referral traffic from ChatGPT, Perplexity, and Gemini tracked in GA4

How do you track AI referral traffic in GA4? By default, you cannot see it as its own channel. Visits from ChatGPT, Perplexity, Gemini, and Copilot get lumped into Referral or Unassigned, so the fastest-growing source of high-intent B2B traffic stays invisible. The fix takes about ten minutes: create a custom channel group with a source regex matching chatgpt|perplexity|claude|gemini|copilot, then report on it like any other channel.

That setup is the easy half. The harder half is what this guide is really about: AI-referred visitors convert at multiples of organic search visitors, most of them arrive anonymous, and GA4 alone will never tell you which target accounts they represent. If you want to see whose companies are behind your AI referrals right now, Book a demo and we will show you on your own traffic.

Why AI referral traffic is invisible in default GA4

GA4 assigns every session to a default channel group: Organic Search, Direct, Referral, Paid Search, and so on. Those rules were written before AI assistants sent meaningful traffic, so there is no "AI" channel. When someone clicks a citation in a ChatGPT answer, GA4 sees a referral from chatgpt.com and files it under generic Referral, next to links from partner blogs and directory sites.

It gets worse. Some assistant clicks pass no referrer at all, especially from native mobile and desktop apps. Those sessions land in Direct, indistinguishable from someone typing your URL. And Google AI Overviews clicks are bundled into Organic Search with no separate label, because Google treats them as search clicks.

The result is a reporting blind spot at exactly the wrong moment. Multiple 2026 analyses have measured AI referral traffic growing at triple-digit rates year over year, with one widely cited study putting growth above 500 percent. If your channel report cannot isolate it, you are making budget decisions with a category-sized hole in the data. We cover how buyers actually behave inside these tools in our guide to how B2B buyers research vendors in AI search.


Step-by-step: build an AI channel in GA4 with a custom channel group

GA4 custom channel groups let you define your own classification rules without touching the default group, so nothing breaks in existing reports. Here is the exact setup we run.

Step 1: Create the custom channel group

  1. In GA4, go to Admin, then under Data display choose Channel groups.
  2. Click Create new channel group. GA4 copies the default group as your starting point.
  3. Name it something explicit, like Channels + AI, so teammates know it differs from the default.

Step 2: Add the AI referral channel with a source regex

Inside the new group, click Add new channel and name it AI Referrals. Set the condition to Source matches regex and use a pattern that covers the assistants that matter today:

.*chatgpt.*|.*openai.*|.*perplexity.*|.*claude.*|.*gemini.*|.*copilot.*|.*edgeservices.*

This catches chatgpt.com, chat.openai.com, perplexity.ai, claude.ai, gemini.google.com, and Microsoft Copilot referrers including the Edge sidebar. Two important details. First, drag the AI Referrals channel to the top of the rule order, because GA4 assigns each session to the first matching channel and Referral would otherwise claim it. Second, revisit the regex quarterly: new assistants and new referrer domains appear constantly, and Google documents channel group behavior in its official Analytics help pages if you need the matching rules.

Step 3: Report on it and mark the conversions

Custom channel groups apply from creation onward for attribution, though GA4 can retroactively classify session traffic dimensions in many standard reports. Open any report with a channel dimension, switch the dimension to your new group, and AI Referrals appears as its own row. From there, compare it on the metrics that matter: engaged sessions, key events (demo requests, signups, pricing views), and session key event rate.

Two additions make the data far more useful:

  • Build an Exploration with Session source / medium filtered to your AI regex, so you can see landing pages. AI assistants disproportionately cite deep, specific pages, not homepages.
  • Create a comparison or audience for AI-referred users, so you can watch their behavior across the full journey rather than a single session.

What the numbers say: AI referrals convert like warm intros, not cold traffic

The volume of AI referral traffic is still small for most B2B sites, typically low single-digit percentages of sessions. Teams look at that number and deprioritize it. That is a mistake, because every credible measurement shows the quality is exceptional.

Ahrefs published one of the cleanest datasets: AI assistants drove only about 0.5 percent of their site visitors but 12.1 percent of their signups, meaning an AI-referred visitor was roughly 23 times more likely to convert than the average visitor. Other 2026 analyses report signup conversion rates around 15.9 percent for ChatGPT-referred traffic and 10.5 percent for Perplexity-referred traffic, against low single digits for organic search.

The mechanism is not mysterious. By the time someone clicks through from an AI answer, the assistant has already done the comparison shopping for them. They asked a specific question, got a synthesized recommendation, and clicked your site as a finalist, not as one of ten blue links. They arrive pre-qualified and late-stage.

If your AI Referrals channel shows even 1 to 2 percent of sessions, treat it as a tier-one conversion channel; if it shows near zero, your problem is upstream visibility, not analytics, and you should start with measuring your AI search visibility before optimizing conversion.


Skip the manual work

Abmatic AI runs targets, sequences, ads, meetings, and attribution autonomously. One platform replaces 9 tools.

See the demo →

The measurement gaps GA4 cannot close

Even a perfect channel group undercounts AI influence. Be honest about these gaps when you report the numbers internally.

Unlinked mentions and zero-click answers

Most AI answers that mention your brand include no clickable citation at all, and many users never click even when links exist. The assistant answers the question, the buyer forms an opinion, and no session is ever created. This is the same zero-click dynamic reshaping search generally, which we break down in our B2B website strategy for zero-click AI search.

Dark AI traffic

Referrer stripping from native apps, privacy features, and some in-answer browsing flows means a share of genuine AI-referred sessions lands in Direct. Your AI Referrals channel is a floor, never a ceiling. A practical heuristic: if Direct traffic to deep technical pages rises alongside your measured AI referrals, dark AI traffic is the likely cause.

Buyers frequently research in ChatGPT, then Google your brand name to visit. GA4 records branded organic or Direct. The AI touch that created the demand is invisible, which is why citation tracking and referral tracking are complementary layers, not substitutes.


From measurement to conversion: what to do with AI-referred visitors

Here is the strategic punchline: knowing that AI referrals convert 10 to 23 times better only matters if you act on individual sessions. GA4 tells you a channel performed well last month. It will never tell you that someone from a target account landed from Perplexity five minutes ago. That is where an account-based platform earns its keep.

Abmatic AI closes that gap in three moves:

  • Identify who is behind the session. Abmatic AI deanonymizes traffic at both the account level (Demandbase and 6sense class) and the contact level (RB2B and Warmly class), natively. An AI-referred session stops being an anonymous GA4 row and becomes a named company, and often a named person, with firmographics attached.
  • Personalize for a pre-qualified visitor. Because AI-referred visitors arrive late-stage, generic pages waste them. Abmatic AI's web personalization (Mutiny class) can serve industry-specific proof, relevant case studies, and a banner CTA gated on the visitor matching your ICP, with A/B testing (VWO class) to validate the lift.
  • Route target accounts into plays automatically. Agentic Workflows can watch for the combination of AI referral source plus target-account match, then enroll the account in an outbound sequence, alert the owning AE in Slack, and launch retargeting across LinkedIn Ads and Meta Ads. Agentic Chat can engage the visitor live, already knowing the account and its intent history.

The final piece is CRM hygiene. Sync the AI referral source into Salesforce or HubSpot via Abmatic AI's bi-directional integration and stamp opportunities as AI-influenced. Within a quarter you can answer the question every CFO is starting to ask: how much pipeline did AI search actually create? Teams doing this today are consistently surprised by the answer. See it live on your own traffic.


How this fits your broader AI search measurement stack

Think of AI search measurement as two layers. Pre-click: how often assistants cite and recommend you, which we cover in how to measure AI search visibility. Post-click: what cited visitors do on your site and in your pipeline, which is this guide. You need both, because optimizing citations without conversion tracking is vanity, and conversion tracking without citation visibility leaves you unable to explain why the channel grows or shrinks.

A quarterly operating rhythm that works: review citation share of voice, review the GA4 AI Referrals channel for volume and key event rate, review identified-account sessions and AI-influenced pipeline in your CRM, then feed what you learn back into content strategy. If a page earns citations but converts poorly, fix the page experience; if a page converts well but earns no citations, fix its answer-engine visibility.

FAQ

How do I track ChatGPT traffic in GA4?

Create a custom channel group in Admin under Channel groups, add a channel named AI Referrals, and set Source matches regex to a pattern like .*chatgpt.*|.*openai.*. Move it to the top of the rule order. ChatGPT sessions then appear as their own channel row in any report using that group.

What regex should I use for AI referral traffic?

A solid starting pattern is .*chatgpt.*|.*openai.*|.*perplexity.*|.*claude.*|.*gemini.*|.*copilot.*|.*edgeservices.*. It covers ChatGPT, Perplexity, Claude, Gemini, and Microsoft Copilot referrer domains. Review it quarterly, because assistants add new domains and new products regularly.

Why does ChatGPT traffic show up as Direct in GA4?

Clicks from native ChatGPT apps and some privacy-restricted contexts pass no referrer header, so GA4 cannot attribute the source and defaults the session to Direct. This dark AI traffic means your measured AI referral channel is an undercount, a floor rather than the true total.

Does Google AI Overviews traffic count as AI referral traffic?

No. Google bundles AI Overviews clicks into Organic Search in both GA4 and Search Console, with no separate label. Your custom AI channel captures assistant referrers like ChatGPT, Perplexity, Gemini, and Copilot, but AI Overviews influence stays hidden inside organic metrics.

Substantially better in every credible study. Ahrefs measured AI visitors at 0.5 percent of traffic but 12.1 percent of signups, roughly a 23x conversion multiplier. Other 2026 analyses report around 15.9 percent conversion for ChatGPT referrals and 10.5 percent for Perplexity, several times typical organic rates.

Can I see which companies are visiting from ChatGPT or Perplexity?

Not with GA4 alone, which reports anonymous aggregates. An account identification platform like Abmatic AI deanonymizes AI-referred sessions at the account and contact level, matches them against your target account list, and can trigger personalization, chat, and sales alerts for the ones that fit your ICP.

Should I invest in AI referral traffic if it is only 1 percent of sessions?

Yes. Low volume with 10x to 23x conversion economics makes it your most efficient channel per session. Small percentages of traffic can drive double-digit percentages of signups and pipeline, and the volume trend is steeply upward, so building measurement and conversion muscle now compounds.

Ready to turn AI referrals into pipeline instead of a GA4 row? Abmatic AI identifies the companies and contacts behind those sessions, personalizes what they see, and routes the right accounts to your sales team automatically. Book a demo.

Run ABM end-to-end on one platform.

Targets, sequences, ads, meeting routing, attribution. Abmatic AI runs all of it under one login. Skip the 9-tool stack.

Book a 30-min demo →
[ KEEP READING ] / related posts
B2B revenue team reviewing why AI SDR booked meetings fail to convert into closed pipeline

Why AI SDR Meetings Don't Convert: The 2026 Reckoning

Fintech marketing team scoring ABM agency proposals during a vendor selection review

How to Hire an ABM Agency for Fintech: Vetting Questions, Red Flags, and the In-House Alternative

Marketing team grouping customers into segments on a whiteboard during a customer segmentation planning session

Customer Segmentation: The Complete Guide (Types, Models, and How to Do It)