How to track ChatGPT traffic in your website analytics
Visitors from ChatGPT, Perplexity, Claude and Gemini often end up filed as "Direct" or generic referral traffic. Here is how to recognise them, measure them properly, and decide what to do about them.
More and more people ask an AI assistant before they search. When ChatGPT, Perplexity, Claude, Gemini or Copilot recommends a page and includes a link, some of those people click through to your website. Those visitors are often unusually well informed: they have already read a summary, compared options and decided your page is worth a look.
The problem is that most analytics setups do not show them as a group. They get scattered across "Referral", lumped into "Direct", or hidden under a long tail of unfamiliar domains. This guide explains why that happens and how to measure AI traffic properly.
Why AI visits are hard to see
A visit is normally attributed using two signals: the referrer (the page the visitor came from, sent by the browser) and campaign tags in the URL (such as utm_source).
AI assistants send both signals inconsistently:
- Web apps usually pass a referrer. A click from the browser version of an assistant normally arrives with a referrer such as
chatgpt.com,perplexity.ai,claude.ai,gemini.google.comorcopilot.microsoft.com. - Desktop and mobile apps often do not. When a link opens from a native app, the browser may receive no referrer at all, so the visit looks like someone typed your address. That is why part of your AI traffic hides in "Direct".
- Some assistants tag their links. ChatGPT adds
utm_source=chatgpt.comto many of the links it shows in answers, which identifies the visit even when the referrer is missing.
Most analytics tools then file whatever they do recognise under a generic "Referral" channel, next to blogs and directories. The total exists, but nobody looks at it.
Step 1: list the sources that count as AI
Start with a list of the domains and tags that should count as AI traffic. A practical starting set:
| Assistant | Referrer domains | Common tag |
|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com | utm_source=chatgpt.com |
| Perplexity | perplexity.ai | |
| Claude | claude.ai | |
| Gemini | gemini.google.com | |
| Copilot | copilot.microsoft.com | |
| DeepSeek | chat.deepseek.com | |
| Grok | grok.com | |
| Le Chat | chat.mistral.ai |
Keep the list up to date. New assistants appear regularly, and existing ones change their domains.
Step 2: give AI traffic its own channel
Once you know which sources count, group them into a dedicated AI channel instead of leaving them in Referral. In most tools that means creating a custom channel group or a segment with a rule such as "source matches any of the domains above, or utm_source is chatgpt.com".
Two details matter:
- Match on both the referrer and the tag. Matching only on the referrer misses the app visits that carry a tag but no referrer.
- Put the AI rule before the Referral rule. Channel rules are usually evaluated in order. If Referral comes first, it swallows the AI visits.
Step 3: compare AI visitors with the rest
Counting visits is only the start. The useful questions are comparative:
- Do AI visitors convert better or worse than search visitors? They often arrive with more context, so a higher conversion rate is common, but measure it on your own site before acting on it.
- Which pages do assistants send people to? These are the pages AI tools consider the best answer to some question. They deserve a clear next step: a product link, a signup, a demo.
- Is the trend growing? A week-over-week view per assistant shows whether a single mention caused a spike or whether you are steadily becoming a source.
Step 4: look at the crawlers too
Referral clicks only show the visitors who clicked. Before an assistant can recommend a page, its crawler usually has to read it, and those crawlers rarely run JavaScript, so script-based analytics never sees them. Your web server's access log does. Our article on AI crawlers explains which bots to look for and how to read the log.
What we do automatically
In our analytics, visits from AI assistants form their own AI channel from the first day. Recognition uses both the referrer and the utm_source tag, the list of assistants is built in, and the dashboard compares AI visitors' conversion rate with your other channels. When an assistant starts sending noticeably more or fewer visitors, it shows up in your daily brief with the numbers behind it.
If you would like to see your own AI traffic, you can start a free trial and install the script in a few minutes, or read more about AI traffic analytics.