Last updated: 17 October 2026
ChatGPT and Perplexity traffic appears in GA4 (Google Analytics 4) primarily under the "Direct" or "Unassigned" channel unless you build a custom filter, because GA4's default channel rules were not designed to recognise AI referrers. UK sites remain especially exposed: ChatGPT drove 86.27% of UK AI referrals, according to SE Ranking (via Tom Riley) (2026), making correct setup a measurable reporting gap.
Key Takeaways
- GA4 (Google Analytics 4) only reliably separates AI referral traffic once a custom channel group or regex filter is built, because the platform's default rules were written before AI chatbots sent meaningful referral volume.
- ChatGPT sent 79.4% of AI chatbot referrals worldwide in August 2026, against Perplexity's 4.3%, according to Statcounter (via Upgrowth) (2026).
- UK websites are more ChatGPT-dependent than the global average, with ChatGPT holding 86.27% of UK AI referral share versus a global figure closer to 75-79%, according to SE Ranking (via Tom Riley) (2026).
- AI referral traffic grew 16-fold in under two years, reaching 0.32% of total website traffic in 2026 up from 0.02% in 2026, according to SE Ranking (2026).
- Aether AI's own publishing operation, which produced 281 articles across four brands in the last 30 days, uses this exact GA4 configuration to attribute AI-driven sessions back to specific published pages.
What Does GA4's Default Classification of ChatGPT and Perplexity Traffic Look Like?
GA4's default channel grouping is Google's built-in system for bucketing sessions into categories such as Organic Search, Referral, Direct and Paid Search, based on the referrer URL and UTM parameters present on arrival. ChatGPT and Perplexity traffic does not fit neatly into any of these legacy buckets, so it gets scattered across several of them depending on how the link was generated.
When a user clicks a citation link inside a ChatGPT answer, the referrer header is often chatgpt.com or chat.openai.com, and GA4 generally classifies this as Referral, provided the referrer isn't stripped. Google introduced a dedicated "AI Assistant" channel grouping in GA4 during 2026 to capture some of this traffic automatically, but coverage is inconsistent across chatbot apps, mobile wrappers and in-app browsers.
Perplexity presents a harder problem. Its answers frequently open citation links in a way that loses the referrer entirely, which means the session lands in GA4 as "Direct" traffic — a user who apparently typed your URL straight into their browser, when in fact an AI engine sent them. This referrer-stripping is the single biggest reason AI-driven sessions get hidden inside Direct and Unassigned, understating true AI influence on a site's traffic.
Which Referrer Domains and URL Parameters Should You Filter For?
Capturing ChatGPT and Perplexity traffic reliably means matching on a specific list of referrer domains and query parameters inside a GA4 custom channel or Exploration filter. The core domains to include in a regex match are chat.openai.com, chatgpt.com, perplexity.ai and www.perplexity.ai, alongside Google's own AI surfaces where relevant.
A workable regex pattern for a GA4 session source/medium filter looks like this:
chatgpt\.com|chat\.openai\.com|perplexity\.ai
Beyond the referrer domain, watch for UTM parameters some AI tools now append to outbound links — typically utm_source=chatgpt.com or utm_source=openai. Build your filter to catch both the raw referrer and any UTM-tagged variant, because a single AI platform can send traffic through more than one route depending on whether the user is on web, iOS, Android or a browser plugin.
It's worth noting that Google AI Overviews traffic, Gemini traffic and Microsoft Copilot traffic each need their own entries in the same regex if you want one unified view — Aether AI's own citation tracking spans exactly these six engines (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot) because no single referrer pattern captures all of them.
How Do You Create a Custom Channel Group to Isolate AI Referral Traffic?
A custom channel group in GA4 is a user-defined set of rules that overrides or supplements Google's default channel grouping, letting you label sessions matching your regex as a named channel such as "AI Referral" rather than letting them fall into Referral, Direct or Unassigned. This is the single most durable fix for the classification problem described above.
To build one, navigate to Admin > Data display > Channel groups within your GA4 property, then select Create custom channel group. Add a new channel named something clear like "AI Assistants", and set the matching condition to Session source / medium matches regex, pasting in the domain pattern covering ChatGPT and Perplexity.
Order matters: GA4 evaluates channel rules top-down, so place your AI Assistants rule above generic Referral and Organic Search rules, otherwise a ChatGPT session may get claimed by a broader rule first. Save the group, then apply it as a secondary dimension on your acquisition reports going forward — note that custom channel groups only process traffic from the point of creation onward, they do not retroactively reclassify historic sessions.
What Navigation Path Shows This Traffic Inside GA4?
The standard navigation path for viewing this data is Reports > Acquisition > Traffic acquisition, where you can swap the default "Session default channel group" dimension for your new custom "AI Assistants" channel using the dropdown at the top of the table. This report shows sessions, engaged sessions, key events and revenue broken down by channel.
For a more targeted view, go to Reports > Acquisition > Traffic acquisition, then click the pencil icon to add a secondary dimension of Session source/medium. This reveals the raw referrer strings GA4 is capturing, letting you confirm chatgpt.com and perplexity.ai are actually appearing before you trust the custom channel's output.
A quicker sense-check for smaller sites is Reports > Acquisition > User acquisition, filtered by First user source/medium, which isolates new visitors arriving specifically via these referrers rather than blending them with returning-user sessions that may have been reclassified as Direct on a later visit.
How Do You Build a GA4 Exploration to Compare ChatGPT vs Perplexity?
A GA4 Exploration is a free-form reporting tool inside the Explore section that lets you build custom tables and breakdowns beyond the fixed structure of standard reports, and it's the best way to see ChatGPT and Perplexity side by side. Navigate to Explore > Blank to start.
Set the technique to Free form, add Session source/medium as a dimension, and add Sessions, Engaged sessions, Average engagement time and Key events as metrics. Apply a filter where Session source/medium contains regex your AI domain pattern, then drag Session source/medium into the Rows field.
This produces a single table with one row for chatgpt.com / referral and one for perplexity.ai / referral, letting you compare volume and engagement quality directly. Add Landing page as a secondary dimension in the Rows field to see exactly which articles or product pages each AI engine is sending visitors to — critical for understanding which content is actually earning citations.
| Comparison point | ChatGPT | Perplexity |
|---|---|---|
| Global AI chatbot referral share, Aug 2026 | 79.4% | 4.3% |
| UK AI referral share (year-on-year) | 86.27% (from 86.88%) | Smaller, declining share |
| Global trend (Apr 2026 to Aug 2026) | Growing dominance | Fell from 12.1% to 4.3% |
| Typical GA4 default classification | Referral or AI Assistant channel | Often Direct (referrer stripped) |
Source: SE Ranking (via Tom Riley) (2026); Statcounter (via Upgrowth) (2026)
Are There UK GDPR or PECR Consent Considerations Affecting This Tracking?
PECR (the Privacy and Electronic Communications Regulations) is the UK law governing cookies and similar tracking technologies, and it directly affects whether GA4 can log an AI-referred session at all. If a UK visitor arriving from ChatGPT or Perplexity rejects analytics cookies in your consent banner, that session will not appear in GA4 regardless of how well your channel group is configured.
The ICO (Information Commissioner's Office) has stated plainly that "online advertising purposes are not exempt from PECR's consent requirements and never have been" — a position that extends to analytics cookies used for attribution. In April 2026, the ICO finalised updated guidance on storage and access technologies, clarifying how PECR and UK GDPR apply to tools like GA4.
Separately, Aether AI notes that the Data (Use and Access) Act 2026 introduced a new "statistical purposes" exemption that may reduce consent friction for pure analytics cookies in some configurations, though businesses should treat this as a legal question for their own advisor rather than a blanket licence. Encouragingly, the ICO reported in April 2026 that 99% of the UK's top 1,000 websites now meet its compliance checks for advertising cookies, according to ICO (via Datashyre) (2026), showing the direction of travel toward tighter but clearer compliance.
Practically, document your GA4 AI-tracking configuration alongside your cookie policy, noting which consent category covers GA4 and whether your banner allows granular opt-in for analytics specifically. A rejected consent banner is one of the most overlooked reasons AI referral numbers look lower than reality.
What Mistakes Cause AI Referral Traffic to Be Misattributed in GA4?
Referrer stripping is the most common cause of misattribution, happening when a browser, app wrapper or security setting drops the HTTP referrer header before GA4 can log it, causing a genuine ChatGPT or Perplexity visit to register as Direct traffic. This single issue likely understates AI referral volume across most UK GA4 properties today.
A second common error is channel rule ordering — placing a custom AI Assistants rule below the default Referral or Organic Search rules in the channel group editor, so GA4 claims the session for a broader category before your specific rule ever runs. Always check rule order after any edit.
A third mistake is relying only on session-level data and ignoring Google Search Console, the free tool showing how your pages appear in Google's AI Overviews and AI Mode. GA4 shows what happened after a click; Search Console (and server log analysis of crawler activity) shows AI engines reading your content even when no click ever occurs — a distinction many UK marketing teams miss when reporting "AI visibility" to the board using GA4 alone.
A fourth is forgetting that UTM-tagged links override referrer-based classification, meaning if an AI platform updates its link format, an existing regex can suddenly stop matching and traffic silently reverts to Direct without any error or warning in GA4's interface.
Your GA4 AI-traffic tracking checklist
- Confirm raw referrer strings for
chatgpt.com,chat.openai.comandperplexity.aiare appearing under Session source/medium before building rules. - Create a custom channel group named "AI Assistants" under Admin > Data display > Channel groups, using a regex covering all six major AI engines.
- Order the AI Assistants rule above default Referral and Organic Search rules in the channel group editor.
- Build a GA4 Exploration comparing ChatGPT vs Perplexity by sessions, engaged sessions and landing page.
- Cross-check GA4 session data against Google Search Console's AI Overviews reporting to catch zero-click AI visibility.
- Review your cookie consent banner's analytics category against current ICO guidance and the Data (Use and Access) Act 2026.
- Re-check referrer capture monthly, since AI platforms frequently change link formats without notice.
- Document the finished configuration alongside your cookie policy for audit purposes.
FAQ
Does GA4 automatically track AI traffic from ChatGPT or Perplexity?
GA4 captures some AI referral traffic automatically through its newer "AI Assistant" channel grouping, but coverage is inconsistent, especially for Perplexity where referrer headers are often stripped. Most UK businesses need a custom channel group or regex filter to see the full picture.
Why does Perplexity traffic show up as "Direct" instead of referral?
Perplexity traffic appears as Direct when the referrer header is stripped during the click-through, which happens often with in-app browsers and some mobile link formats. GA4 then has no referrer string to classify, so it defaults the session to Direct rather than Referral.
What UTM parameters does ChatGPT use for referral links?
ChatGPT links sometimes carry utm_source=chatgpt.com or utm_source=openai, though this is not applied consistently across every surface. The safest approach is to build your GA4 filter around both the raw referrer domain and any UTM variant, rather than relying on UTMs alone.
Is consent required for GA4 analytics cookies under UK GDPR and PECR?
Yes, PECR generally requires consent before setting non-essential cookies including GA4's analytics cookies, and the ICO has confirmed this applies to tracking used for measurement, not only advertising. The Data (Use and Access) Act 2026 introduced a narrower "statistical purposes" exemption, but most UK sites still need a compliant consent banner.
How long does it take for new AI referral data to appear and stabilise in GA4?
Standard GA4 reports typically populate within 24-48 hours of a configuration change, though Exploration reports can show data almost in real time. Custom channel groups only apply going forward, so expect a few weeks of data collection before trends look statistically stable.
Is there any cost involved in setting this up in GA4?
No, building a custom channel group, regex filter or Exploration report is entirely free within GA4's standard tier and requires no GA4 360 upgrade. The only cost is analyst or marketer time spent configuring and maintaining the rules.
Who in a UK business should own this GA4 configuration?
Typically a marketing analyst or in-house SEO lead builds and maintains the channel group, with IT or a data team involved where consent management platforms need adjusting. In smaller UK businesses, one marketing generalist often owns the whole setup alongside Google Search Console monitoring.
How does GA4 compare to Search Console or server log analysis for tracking AI visits?
GA4 shows what happened after a visitor clicks through from an AI engine, while Google Search Console shows impression data for how your pages appear inside AI Overviews even without a click. Server log analysis goes a step further, revealing AI crawler activity before any human visit occurs — the three together give a fuller picture than GA4 alone.
Seeing Your Own AI Referral Traffic Clearly With Aether AI
Getting ChatGPT and Perplexity to cite a page is only half the job — confirming it actually drove a visit, inside a GA4 report your team trusts, is the other half. Aether AI was built around exactly this gap, pairing automated, AI-optimised article generation with citation tracking across six engines (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot), so the content work and the measurement work sit in one place rather than two disconnected tools.
Aether AI's own operation is the proof: across its own four brands — spanning security, facilities software and branding — the platform published 281 articles in the last 30 days, with GA4 and Google Search Console integration used to track exactly which of those pages earned AI citations and referral clicks.
If you want to see where your own site currently stands, Aether AI offers a free AI-visibility audit at /audit, alongside public pricing for teams ready to track citations and traffic properly from day one.