Your AI traffic is bigger than your dashboard says
Most visits from ChatGPT and Perplexity land in your analytics as 'Direct.' Here's why, and how to see the real number.
Open your analytics and look at how much traffic you got from ChatGPT last month. If the number looks tiny, or zero, don't relax. It's almost certainly wrong, and it's wrong in a direction that hides one of the fastest-growing sources of visitors you have.
Most of your AI traffic is sitting in the "Direct" bucket, unlabelled, mixed in with people who typed your URL by hand.
Why AI visits disappear
When someone clicks a link, the browser usually tells the destination where they came from. That referrer is how analytics tools know a visit came from Google or Twitter or a newsletter. AI assistants break this in a few ordinary ways.
A lot of AI answers are read inside an app or a mobile webview that never sends a referrer. Some platforms strip it on purpose for privacy. Links rendered with rel="noreferrer" send nothing by definition. And until recently, several assistants didn't make their citations clickable at all, so the visit arrived through a copy-paste with no trail.
The result is consistent across the industry: a large majority of real AI-referred sessions show up with no referrer and get filed as Direct. Published estimates put it around 60 to 70 percent. So the ChatGPT line in your dashboard isn't your ChatGPT traffic. It's the small slice that happened to keep its referrer, and you're treating it as the whole.
Why this is the wrong number to get wrong
AI traffic is still small for most sites, somewhere in the low single-digit percentages, higher if you're in tech or ecommerce. Easy to ignore at that size. Two reasons not to.
It converts. These are people who asked an assistant a question, got your site recommended as the answer, and arrived already half-sold. Undercounting your best-converting channel is an expensive mistake to make quietly.
It's growing faster than anything else you have. A channel that's small, high-converting, and growing that fast is exactly the one you want an accurate read on now, before it's the channel that matters most.
How to actually see it
The fix is to stop relying only on the referrer. If you can fingerprint the session and match the signals an AI visit actually carries, you can pull a big chunk of that "Direct" pile back into the AI column where it belongs:
- ✓The UTM tags some assistants append to their links
- ✓Landing-page and session patterns typical of AI answers
- ✓Known AI sources, including the ones GA4 skips
- ✕Referrer headers alone, most AI visits don't carry one
That's what usrPeek does. It classifies AI referrers including the ones GA4 misses, like Perplexity and Copilot, and it surfaces the AI visits that would otherwise vanish into Direct. You see how many people each assistant actually sent you, what they did on the page, and whether they clicked your CTA or converted. Not a referrer fragment. The real source-to-outcome story.
There's a second half most analytics can't touch at all, which is AI crawlers. The bots that read your site to train and answer (GPTBot, ClaudeBot, PerplexityBot) don't run JavaScript, so a client-side tool literally cannot see them. usrPeek tracks them server-side, so you can also tell whether the assistants are even reading you yet. That's a separate post.
The ChatGPT number in your dashboard is a floor, not a fact.
For now, the takeaway is small and worth acting on. Point usrPeek at your site and find out how much AI traffic you've actually been getting credit for none of.
See what your visitors actually do
usrPeek is dead-simple analytics for solo founders. One script tag for visitor journeys, auto-tracked clicks, AI traffic, and a daily email. No bolt-on cookie tooling.
Start for free