Why GA4 hides it by default

GA4's channel definitions were written before AI search existed, so there is no built-in "AI" bucket. When ChatGPT or Perplexity does send a referrer, GA4 usually files it under Referral, mixed in with every other site. When the assistant sends no referrer - common on mobile apps and on Google's own AI surfaces - the visit has no source to read, so GA4 calls it Direct, the same bucket as someone typing your URL. The result is that your AI traffic is real but smeared across three channels, with the largest piece hiding inside Direct where you would never think to look for it.

The fix: a custom channel group

In GA4, go to Admin, then Channel groups, and create a custom group with a new channel - call it AI - defined by the referrer or source matching the assistant domains. A regex like this captures the major ones:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|meta\.ai

One piece of good news: ChatGPT began appending utm_source=chatgpt.com to its citation links in 2025, so those clicks tag themselves and are easy to isolate even without the channel group. Two pieces of fine print: the channel group does not backfill, so it only sorts traffic from the day you create it onward, and the domain list drifts as products get renamed and new engines launch, so revisit it every few months.

Why it is still an undercount

Even a clean channel group only catches the visits that arrive with a readable source. A large share of AI-driven traffic carries no referrer - from assistant mobile apps, from Google's AI Overviews and AI Mode, from copy-pasted answers - and those land in Direct with no way to reclassify them. So the figure you build is a confirmed minimum, not the true total. That is fine, as long as you read it that way: a floor that tells you AI is sending at least this much, trending in some direction, rather than a precise count you can take to the bank.

Traffic is not the whole story

The bigger gap is one no GA4 setting can fix. Analytics can only see a click. But AI answers routinely resolve a question without sending anyone anywhere - the buyer reads that you are the right choice, forms an impression, and never lands on your site. That citation shaped the decision and left no trace in GA4. So if you judge your AI visibility purely by referral traffic, you will undercount your real influence and, worse, you will not even know when an engine is describing you wrongly to people who never click.

The honest approach runs two meters at once: referral traffic for the clicks AI sends, and citation tracking for the far larger set of answers that mention you with no click. The second is the subject of measuring AI citations, and the hands-on version is how to check if AI search mentions your brand. If your traffic looks thin, the question underneath it is usually whether you are being cited at all.

How Web Cited helps

GA4 tells you how many people AI sent to your door. It cannot tell you how often AI named you to people who stayed put, or whether it got you right. That is the half the SXO Audit measures: it runs 25 buyer prompts across six engines with three trials each over time and reports your citation share, so you can pair the clicks GA4 sees with the mentions it never will. Start with the free 10-minute AI search audit or the Free Snapshot for a first read.

Try the Free Snapshot   See the SXO Audit

By the Web Cited Editorial Research Team. Last updated 1 June 2026.