crontent

Most of your AI traffic is hiding in GA4

Crontent

Two founders can publish the same GEO playbook this month and get the same traffic bump. One will keep going. The other will stop in two weeks. The difference is simple: one can prove whether AI visits turn into signups.

How do I prove which AI traffic converts?

Start by separating confirmed AI traffic from AI-influenced traffic. Swetrix says to “prove the signals you can observe, label the signals you infer, and keep both groups apart in reports.” That's the right frame.

Confirmed AI traffic is the easy part. Someone clicks from chatgpt.com, perplexity.ai, claude.ai, gemini.google.com, or copilot.microsoft.com, and you can capture that referrer. Tagged AI links count too, if you control the link and add UTMs.

AI-influenced traffic is messier. Someone reads about you in ChatGPT, then comes back later through Google brand search or direct. You can't prove all of that from analytics alone. So don't pretend you can. Keep a separate bucket for things like:

  • direct visits to pages AI tools cite
  • Google clicks after AI Overviews or AI Mode
  • demo forms where people type “found you in ChatGPT”
  • CRM notes from sales calls

That split matters because it stops you from lying to yourself. If you lump confirmed and suspected AI traffic together, every report turns into vibes. If you keep them apart, you can answer the only question that matters: which visits actually make money?

GA4 hides enough AI traffic to wreck your decisions

GA4 can bury AI visits inside Direct, Organic Search, and Unassigned. Flux says default channel logic misses 60-80% of AI referrals because in-app browsers strip referrers, some AI clicks pass through redirect chains, and GA4's default grouping was not built for AI answer engines.

That means your dashboard can say “Direct is up” when what really happened is ChatGPT sent you a buyer.

Metrbox shows the practical fix. Use a GTM classifier that checks UTM sources, browser referrers, and optional custom host mappings, then writes a clean label like chatgpt, perplexity, claude, gemini, or copilot into GA4 as an event parameter and custom dimension such as ai_source.

You do not need a perfect attribution model to get value here. You need a cleaner field than GA4 gives you out of the box.

If you're a tiny SaaS team, that changes what you do next week. You stop asking “are we getting AI traffic?” and start asking:

  • which AI source sent the visit
  • which landing page got cited
  • which of those visits signed up
  • which signups turned into revenue

That is already more useful than another month of posting into the void.

The GEO edge is a feedback loop, not more output

The useful moat right now is not better prompting. It's being one of the few teams that can connect AI citations to sessions, signup rate, and revenue.

Swetrix lays out a simple reporting map: segment confirmed AI referrers by landing page, event, goal, funnel, and revenue; compare tagged AI links against referrer revenue; pair Google Search Console with landing page and revenue data for Google AI clicks; and mark suspected AI direct separately.

Once you have that loop, budget decisions get easier fast.

If Perplexity traffic signs up well on comparison pages, make more of those. If ChatGPT sends traffic to your tutorial pages but none of it activates, fix the page or stop chasing that format. If AI traffic is low volume but high intent, you don't need a giant content calendar. You need tighter pages, clearer proof, and a stronger call to action.

That also changes hiring. A lot of small teams think the next move is a full-time content person. Usually it isn't. The better first hire is someone who can wire tracking, test pages, and read the conversion data without making excuses for it.

If you can't tie AI visits to revenue, you're still guessing

AI discovery can look exciting long before it becomes useful. A spike in sessions feels good. It does not tell you where to spend.

Flux says you can build the custom channel group and regex filter in about 30 minutes. Metrbox shows a clean way to pass known AI source labels into GA4. Swetrix gives the reporting rule that keeps you honest: prove what you can observe, infer the rest, and don't mix them.

So before you write one more GEO post, fix measurement.

Track visible AI referrers. Tag the links you control. Mark suspected AI influence without pretending it's confirmed. Then look at signups and revenue, not screenshots of traffic graphs.

The teams that win this round won't be the ones publishing the most. They'll be the ones who can say, with receipts, which AI traffic actually pays.

Sources