Measuring AI Traffic in Your Analytics

AI traffic is already in your analytics, unlabeled, filed under referrals nobody looks at. Twenty minutes of setup makes it a channel you can read: how big, how fast it grows, what it converts, and which of your pages the assistants actually cite.

Here is the segment, the three gauges and the pitfalls that undercount it.

Step 1: build the segment

One filter captures the family: traffic whose referrer or source matches the assistant list.

The one-line version for any tool that takes patterns: match source or referrer against chatgpt|perplexity|gemini|copilot|claude. In GA4, the tidy implementation is a custom channel group with that condition; the quick one is the same regex in an exploration. Either way, save it, because a segment you rebuild monthly is a segment you abandon.

Guard against double counting: tagged ChatGPT clicks carry both the referrer and the utm, so define the segment as one OR condition, not two stacked filters.

Step 2: read the value gauge, not the volume gauge

The segment will be small, low single digits of traffic for most sites, and the wrong conclusion is "ignore it". The two numbers that matter are its growth rate month over month and its conversion rate against your site average.

AI-referred visitors arrive pre-advised, as the traffic mechanics explain, and segments like this routinely convert above the site norm. A small channel that converts well and compounds monthly is a channel worth feeding early.

Step 3: watch brand-search lift beside it

The invisible half of AI influence lands as brand search: recommended, remembered, Googled later. The gauge is in Search Console: your brand-query impressions, trended monthly.

Gauge 3 measures from the answer side, prompt tracking through the visibility tools, and its role in the trio is diagnostic: recommendation share rising while referrals stay flat means you are being named without being linked, which points the work at citations.

Step 4: mine the landing-page report

Inside the AI segment, the landing-page dimension is the most actionable report you own: it lists exactly which pages assistants cite, in proportion to how often.

Read it as format feedback. The pages earning AI referrals are usually your comparisons, lists and answer-shaped pieces, and whatever pattern your own list shows is the pattern to publish more of. It also flags the inverse: important pages absent from the list are candidates for a quotability rewrite.

The undercounting pitfalls

Know what the segment misses so you never present it as the whole. Some in-app browsers strip referrers, sending AI clicks into direct. Some assistants link without tags. And the recommendation-without-click path never touches your site at all, which is gauge 2's whole job.

The honest framing for any report: "visible AI referrals, a floor, growing at X percent monthly, converting at Y times site average", with brand lift beside it as the shadow measure.

The monthly ten minutes

Segment trend, conversion against average, top cited pages, brand-query trend: four glances, ten minutes, once a month, filed with the rest of your reporting. The channel is young enough that this cadence catches every meaningful change, and the leaderboard patterns give you the external benchmark when you want one.