Measuring AI search traffic in GA4, step by step

Key takeaways

• AI clicks arrive as referrals from chatgpt.com, perplexity.ai, and friends — GA4 hides them under Referral by default

• A custom channel group with a source regex surfaces them; order it above Referral or it stays empty

• The number is a floor: AI Overviews clicks hide in organic, app sessions land in direct

• Published studies from Microsoft Clarity and Seer Interactive consistently find AI referrals converting well above site average

• The landing-page exploration doubles as a citation report — it shows what AI engines consider your quotable pages

The referrer domains AI platforms send

Each assistant that shows clickable citations passes its own referrer. The ones worth matching today: chatgpt.com (with chat.openai.com still appearing from older surfaces), perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai, meta.ai, and chat.deepseek.com, plus grok.com. ChatGPT also frequently appends utm_source=chatgpt.com to outbound links, which gives you a second, more reliable match key when the referrer is stripped. The list drifts as platforms rename domains — ChatGPT’s own move from chat.openai.com to chatgpt.com re-bucketed everyone’s data mid-stream — so treat the regex as something you revisit quarterly, not something you set once.

• Core set: chatgpt.com, chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai

• Growing set: meta.ai, chat.deepseek.com, grok.com

• ChatGPT often adds utm_source=chatgpt.com — match source as well as referrer

• Revisit the list quarterly; platform domain changes silently re-bucket your data

Build the custom channel group

In GA4, go to Admin, then under Data display choose Channel groups, and create a new channel group rather than editing the default. Add a channel — call it AI Traffic — with the condition that session source matches a regex such as chatgpt|openai|perplexity|gemini|copilot|claude|deepseek|grok, scoped tightly enough to avoid false matches. Order matters: GA4 assigns each session to the first channel whose rule it matches from the top, so drag AI Traffic above Referral and Organic Search, or those channels will capture the sessions first and your new channel will stay empty. Custom channel groups are evaluated at reporting time, so the grouping applies to historical data as well.

• Admin → Data display → Channel groups → create a new group, keep the default intact

• Condition: session source matches your AI domain regex

• Drag the AI channel above Referral — first match wins, top to bottom

• Grouping applies retroactively in reports, so you get history on day one

Build an exploration for the detail

The channel group gives you the headline; an exploration gives you the diagnosis. Create a free-form exploration with session source as a row dimension, filtered by your AI regex, and add sessions, engaged sessions, key events, and revenue as metrics. Add landing page as a second dimension and you can see exactly which pages each assistant cites — which is the report that connects measurement back to content strategy, since it tells you what ChatGPT and Perplexity currently consider your citable pages. Segmenting by platform matters too: assistants differ in what they cite and in the intent of the users they send.

What the numbers systematically miss

Treat GA4’s AI number as a floor, not a total. Clicks from Google’s AI Overviews and AI Mode carry a google referrer, so they land in organic search with no way to separate them in GA4 — Search Console shows them within overall search performance rather than as a distinct AI segment. Sessions from native mobile apps and some in-product browsers arrive with no referrer and fall into direct. Some assistants answer without sending anyone at all, influencing a purchase that later shows up as direct or branded search. The honest reading: the AI channel understates AI influence, and the direction of the error is known.

• AI Overviews and AI Mode clicks hide inside google / organic — not separable in GA4

• App-based assistant sessions often arrive referrer-less and land in direct

• AI-influenced buyers frequently return later via brand search or direct

• Rising direct plus rising branded search alongside AI citations is a telltale pattern

Why AI traffic tends to convert better

Across published analyses the direction is consistent: AI referral traffic converts at a meaningfully higher rate than the site average. Microsoft Clarity’s research put AI traffic conversion at roughly three times other channels, and Seer Interactive’s client analysis found ChatGPT and Perplexity referrals converting at several times the rate of Google organic. The magnitudes vary by study and industry, so benchmark your own property rather than borrowing a multiplier. The mechanism is intuitive: a visitor arriving from an AI answer has already had the product explained, compared, and recommended — the assistant did the top of the funnel, and the click is closer to a decision than a discovery.

From measurement to action

The point of the setup is the loop it enables: the landing-page report shows which pages AI engines cite, conversion by platform shows which citations are worth money, and both feed what you optimize next — more of the guide and comparison content that earns citations, fixes for cited pages that underconvert. The measurement itself is exactly the kind of standing work that suits delegation: WorkDaddy’s Analyst agent maintains the channel definitions and explorations in GA4 and reports the trend, while the Visibility agent works the other side — earning more of the citations the reports measure. Either way, run the setup this week; the history becomes valuable the moment the channel exists.