Updated 4 October 2026

Track ChatGPT and AI traffic

Two different things hide under "AI traffic": people clicking through from an assistant, and robots reading your pages. Here is how to see both, and what each number means.

In short

  • Visitors from AI assistants arrive with a referrer. trckable sorts them into their own channel, AI assistants, with no setup.
  • AI crawlers are robots. They never run JavaScript, so a browser script cannot see them. Your server or CDN reports them with one POST to /api/crawl.
  • The AI & Search tab puts Google, assistants and crawlers side by side, and shows for each page how often AI read it and how many visitors AI sent.

Two kinds of AI traffic#

A visitor from an assistant is a person who clicked a link in an answer. Their browser sends a referrer, and that is all the evidence there is. A crawler is software that fetched your page, either to answer someone right now or to collect training data. It is not a visitor, and mixing the two gives you numbers that mean nothing.

Visitors from assistants#

trckable recognises assistants by their referrer host: ChatGPT, Claude, Perplexity, Gemini, Copilot, DeepSeek, Grok, Meta AI, Mistral, You.com, Phind, Poe and others. Those visits land in the AI assistants channel next to Direct, Search, Social, Email, Paid and Referral. The check runs before the search check, so a visit from Gemini is an assistant and not Google search.

There is nothing to install beyond the normal script. One honest limit: some assistants and apps send no referrer, and those visits show as Direct. The count is a floor, not a total.

Robots reading your pages#

To count crawlers, tell trckable when one asks for a page. Each report is one POST with four fields: your site id, the page address, the robot's user agent and the status.

curl -X POST https://stats.example.com/api/crawl \
  -H 'content-type: application/json' \
  -H 'X-Trckable-Proxy-Key: tkb_px_yourkey' \
  -d '{"s":"tkb_yoursiteid","u":"https://example.com/page","ua":"<user agent>","st":200}'

The docs have copy-ready setups for a Cloudflare Worker, Vercel middleware, nginx and Caddy, and the dashboard's Connect crawler data card fills in your site id and key. Only the robot's name, the page and the day are kept. No address, no cookie, no visitor, and a user agent trckable does not recognise as a robot is thrown away.

Reading the AI & Search tab#

The tab lives in Who came and has three columns:

  • Google: the searches that showed your site and how many were clicked, from your own Search Console.
  • AI assistants: visitors by assistant.
  • AI crawlers: which robots read you, and under them the pages they read most.

Each page row has two numbers, read and sent: "AI read this 400 times, sent 12 visitors." Two badges point at what to look at. No credit means AI crawlers read the page at least 10 times and sent nobody, so it may be used in answers without a link back. Not read means Google sent the page at least 5 clicks and no AI crawler read it in the period.

A click on a row filters the whole dashboard, like any other list.

What to do with it#

A page with many reads and no visitors may need a clearer title or a link worth clicking. A page Google likes and AI never reads may be blocked or hidden behind a script. If you do not want some crawlers, say so in your robots file.

The details, setups and the API call are in the AI & Search docs.

Questions#

Can I see ChatGPT traffic without Google Analytics?
Yes. Visitors who click through from ChatGPT arrive with a referrer, and trckable files them under the AI assistants channel. Nothing extra is needed beyond the normal script.
Why do AI crawlers not show up by themselves?
Robots do not run JavaScript, so a browser script never sees them. Your server or CDN has to report them with one POST to /api/crawl for each robot request.
Is the AI visitor count complete?
No. Some assistants and apps send no referrer, so their visits count as Direct. Treat the number as a minimum.
Start free