How to Tell If AI Search Is Sending You Traffic

AI assistants are quietly sending people to your site. The hard part is proving it. Here's how I actually track AI search traffic without fooling myself.

Why This Is Hard

AI search referrals are badly under-attributed. When someone reads your link inside ChatGPT and clicks it, the click often arrives with a clean referrer, or with no referrer at all.

A few things break the chain:

No referrer header — some assistants strip it, so the visit lands in your analytics as "Direct."
No UTM tags — AI tools don't add campaign parameters. You only get what the platform sends.
App vs. web — clicks from a mobile app or desktop client behave differently than clicks from a browser tab.
Aggregation — analytics tools lump unknown sources into "Direct" or "Other," hiding the signal.

So the honest starting point is this: you will never get a perfect number. You're looking for patterns, not a clean dashboard metric.

Where To Look First

Start with referrer hostnames. When an assistant does pass a referrer, it's usually obvious once you know the names to watch for.

Common ones I filter for:

chat.openai.com and chatgpt.com — ChatGPT
perplexity.ai — Perplexity
gemini.google.com — Gemini
copilot.microsoft.com and bing.com — Copilot and Bing's AI answers

In Google Analytics 4 or Plausible, open your referrers report and search for these strings. Even a handful of sessions confirms the channel is live. Treat each new hostname as a thread to pull.

Then look at your "Direct" traffic. If direct visits to deep, specific pages climb while your homepage stays flat, that's a tell. People don't type long URLs from memory. They got that link somewhere — often from an assistant that dropped the referrer.

Build A Referral Segment

Once you know the hostnames, turn them into a saved filter so you stop checking manually.

The concept is simple. Create a segment that matches when the referrer source contains any assistant hostname:

referrer contains:
  chatgpt.com OR
  chat.openai.com OR
  perplexity.ai OR
  gemini.google.com OR
  copilot.microsoft.com

In GA4, build this as a custom segment or a channel group named "AI Assistants." In Plausible or a log-based tool, it's a single OR filter on the referrer field. Name it clearly so future-you knows what it means.

Now you have one place to watch. Trend it weekly. The absolute number matters less than the direction.

Watch Server Logs And Branded Spikes

Your analytics tag only fires in a real browser with JavaScript. Server logs catch more — including the assistant crawlers that fetch your pages to answer questions in the first place.

In raw access logs, grep your referrer and user-agent fields for the same hostnames plus bot names like GPTBot, PerplexityBot, and Google-Extended. Crawler hits tell you which pages assistants are reading. Referral hits tell you which pages are sending humans back. Both are useful, and they rarely match what your analytics UI shows.

The other signal: branded query spikes. When an assistant cites you by name, people go search for your brand directly. A sudden bump in branded searches or direct homepage visits — with no campaign behind it — often trails an AI mention. It's circumstantial, but combined with referrer data it builds a real picture.

This is the practical side of getting cited by ChatGPT: once you're in the answers, the traffic shows up in fragments across these channels.

The Limits Of Measuring This

Be honest about what you can't see. A lot of AI value never produces a click at all — someone reads your answer inside the chat, trusts it, and never visits. That's still influence. It just doesn't appear in any traffic report.

Referrer data is also inconsistent and changes without notice. A platform that passes referrers today might strip them next month. So don't build a business case on a single number from a single tool.

What works is triangulation: referrer hostnames, plus direct-traffic patterns, plus server logs, plus branded-query movement. No one source is trustworthy alone. Together they tell you whether GEO is actually paying off.

Make It A Habit

Set the segment up once, check it weekly, and watch the trend rather than the noise. Over a few months you'll learn which pages assistants favor and which queries send people back.

If you'd rather not stitch this together by hand, that's the problem I built OptimizeCamp to solve — surfacing where AI search picks you up so you can track AI search traffic and act on it. Start by saving that referral segment today, and let the data accumulate.