Here's a scenario that's happening more often than you'd think. You get a new customer. They sign up, purchase, become a great client. You check your analytics to see where they came from. The source? "Direct." No referral. No campaign tag. No search query. Just... appeared.
There's a good chance that customer was sent by an AI.
So what exactly is AI traffic?
AI traffic refers to website visits that originate from AI platforms like ChatGPT, Perplexity, Claude, Gemini, or Grok. Someone asks an AI assistant a question, the AI recommends your product or links to your site, and the person clicks through to check you out.
It's a real and growing traffic source. But here's the catch: most analytics tools weren't built to track it, which makes it feel like it barely exists even when it's sending you valuable visitors.
Why should you care?
Two reasons. First, AI traffic converts well. Really well.
Think about why. When someone arrives at your site from a Google search, they're in research mode. They might visit five other sites before making a decision. When someone arrives from an AI recommendation, they've already been told "this product is a good fit for what you need." The AI did the comparison shopping for them. They show up pre-sold, or at least pre-interested.
Early data from businesses tracking AI referrals shows conversion rates that significantly outperform traditional organic search traffic. The volume is still smaller, but the quality per visit is often higher.
Second reason: AI traffic is growing fast while traditional organic traffic is flattening for many industries. As more people use AI assistants for product research, recommendations, and purchase decisions, the share of traffic coming from these platforms will keep climbing. Missing this shift means missing where your next customers are actually coming from.
The attribution problem
This is where things get frustrating. Most AI traffic is nearly invisible in Google Analytics, Plausible, or whatever analytics tool you use.
When someone clicks a link in ChatGPT, it doesn't always pass a clean referrer header. Same with Claude, Gemini, and others. The visit often shows up as "direct" traffic, which is the analytics junk drawer for anything without a clear source.
Perplexity is the exception. It tends to pass referral information more reliably, so you might see "perplexity.ai" in your referral sources. But even that isn't consistent.
This creates a real blind spot. You might be getting meaningful traffic from AI platforms and have no idea because your analytics are attributing it to "direct" or "unknown." You can't optimize what you can't measure, and you definitely can't make a business case for investing in AI visibility if you can't show the traffic numbers.
How to get better visibility into AI traffic
There are a few approaches, none of them perfect.
UTM-style tracking through AI citations is limited since you can't control how AI links to you. Some people add specific landing pages for AI-sourced traffic, but this is cumbersome and unreliable.
The more practical approach is to track AI visibility directly. Instead of trying to measure clicks from AI (hard), measure whether AI is recommending you in the first place (achievable). Tools like Mentionable track your brand's presence across AI platforms, showing you which prompts trigger recommendations and which don't. If you're visible in AI answers, the traffic is coming, even if your analytics can't attribute it perfectly.
You can also add a "How did you hear about us?" field to your signup or contact forms. Simple, but surprisingly effective. Many customers will tell you "ChatGPT recommended you" if you just ask.
The honest truth
AI traffic measurement is messy right now. The analytics infrastructure hasn't caught up to the reality of how people discover brands through AI. You'll be operating with imperfect data for a while.
But waiting for perfect tracking before taking AI visibility seriously is like waiting for Google Analytics to be invented before building a website. The traffic is real, it converts, and it's growing. Act on what you can measure, improve tracking as tools mature, and don't let the attribution gap convince you this channel doesn't matter.
