Tracking AI referral traffic from ChatGPT, Perplexity and Claude

Something new shows up in referrer logs since 2024: chatgpt.com, perplexity.ai, claude.ai, gemini.google.com. When an AI assistant cites your page as a source, or a user asks "where can I read more?", people click through, and that click carries a referrer like any other. For sites with strong informational content, AI assistants are quietly becoming a top-five channel.

Why your analytics probably hides it

Most analytics tools classify referrers against lists frozen years ago: search engines, social networks, everything else lumped into "referral". So ChatGPT traffic hides in a generic referral bucket next to some forum from 2011, and Gemini traffic is especially misleading: gemini.google.com pattern-matches "google", so naive classifiers count it as organic search and inflate your SEO numbers.

The domains worth classifying as an explicit AI channel today:

• chatgpt.com and chat.openai.com (ChatGPT)
• perplexity.ai (Perplexity, the most citation-heavy of the group)
• claude.ai (Claude)
• gemini.google.com and bard.google.com (Gemini)
• copilot.microsoft.com (Copilot)
• you.com, phind.com, poe.com, meta.ai, grok.com, deepseek.com

What AI visitors do differently

Early patterns across sites we've measured: AI referrals land deep (an assistant links the exact page that answers the question, not your homepage), they arrive with high intent (someone asked a question and chose to go deeper), and volume is spiky; a single assistant starting to cite one article can double that page's traffic in a week. That spikiness is exactly why AI deserves its own channel: blended into "referral", the trend is invisible until it's huge.

Measuring it

You need three things: referrer-based classification with an AI domain list (see the list above), landing-page breakdowns per channel (to learn which content assistants cite), and trend comparison so a 74% week-over-week AI jump surfaces as an insight instead of noise.

Can you influence it?

Partially. Assistants cite pages that answer specific questions cleanly: clear headings, direct answers near the top, and factual density all help, the same properties that win featured snippets. Crawlability matters too: GPTBot, ClaudeBot and PerplexityBot are well-behaved crawlers you can allow or block in robots.txt. Blocking them keeps your content out of assistants, along with the citations and the traffic. That's a real tradeoff each site has to weigh; measuring the channel first makes it an informed one.