Find ChatGPT visits in GA4's AI Assistant channel, catch AI traffic filed as Direct, check whether OpenAI's crawlers can read your pages, and track whether ChatGPT mentions you.
The short answer
ChatGPT sends you two kinds of traffic, and you track them in different places. People who click a link in a ChatGPT answer are ordinary visits: in Google Analytics 4 they land in the AI Assistant channel, which Google added on 13 May 2026. OpenAI's crawlers (OAI-SearchBot, ChatGPT-User and GPTBot) never run your analytics script, so you only see them in server or edge logs. Neither one tells you whether an answer cited you: for that, use the AI reports Bing and Google publish, or ask a fixed set of questions on a schedule. (Sources checked 6 November 2026.)
| What you want to know | Where to look | |:---|:---| | How many people clicked through from ChatGPT | GA4: Reports > Acquisition > Traffic acquisition, AI Assistant channel, with Session source | | Which pages they landed on and what they did next | GA4: the same report with Landing page, or an Exploration | | Whether OpenAI's crawlers can reach your pages | Server, CDN or edge logs, filtered by user agent | | Whether ChatGPT names or cites you | A fixed prompt set asked on a schedule; Bing Webmaster Tools AI Performance for Copilot |
Step 1: Find ChatGPT visits in GA4
Open Reports, then Acquisition, then Traffic acquisition. In the Default Channel Group you will see AI Assistant as its own row. Google describes it as the channel by which users arrive from sources like ChatGPT, Gemini, Deepseek, Copilot or Grok, and it is matched on the referrer: when the referring site is a recognised AI assistant, GA4 gives the session the medium ai-assistant and the campaign (ai-assistant).
To split the channel by assistant, add Session source as a secondary dimension. chatgpt.com is ChatGPT. To see which pages AI visitors land on, switch the primary dimension to Landing page and filter the channel to AI Assistant.
Two things the channel leaves out on purpose. Clicks from Google's own AI Overviews and AI Mode are Google search, not an AI assistant, so they stay in Organic Search (Search Console counts them in its Web search results). And the channel only covers assistants on Google's list, so check the Session source column for any assistant you expect to see in it.
Step 2: Catch the AI traffic GA4 misfiles
If an AI app opens your link without passing a referrer, GA4 can't tell where the visit came from and files it under Direct. A rise in Direct sessions that land deep in the site, on a specific guide or product page rather than the homepage, is often AI traffic that lost its referrer. Compare Direct landing pages month over month and look for pages you know AI answers point to.
If you want a channel of your own, for example to include an assistant Google doesn't list yet, create a custom channel group in Admin > Data display > Channel groups. Copy the default group, add a channel called AI assistants above Referral, and set its rule to Session source matches regex:
``` ^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|www\.perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|chat\.deepseek\.com|grok\.com)$ ```
Order matters: GA4 assigns a session to the first channel whose rule matches, so the AI channel has to sit above Referral or the visits stay in Referral.
Step 3: See whether OpenAI can read your pages
OpenAI uses three agents, and each does a different job. OAI-SearchBot crawls pages for ChatGPT search. ChatGPT-User fetches a page because a person in ChatGPT asked about it. GPTBot collects pages for training future models. None of them runs JavaScript analytics, so GA4 will never show them.
To see them, search your server or CDN access logs for the user agents. On a typical Linux server: `grep -Ei "OAI-SearchBot|ChatGPT-User|GPTBot" access.log`. A 200 means the crawler got the page. A 403 means something blocked it, often a firewall or bot-protection rule rather than robots.txt. A 429 means it was rate-limited. A run of 404s means it is asking for pages that moved.
Then check robots.txt. With no rule that names them, crawlers are allowed. Blocking GPTBot stops training crawls but doesn't stop OAI-SearchBot, so decide each agent on its own. The free cite-met scan checks the robots.txt side, and on sites cite-met hosts the dashboard charts every AI crawler fetch by page, because every request reaches our edge.
Step 4: Track whether ChatGPT mentions you
Visits and fetches don't prove an answer used your page. A crawler can fetch a page and never quote it, and many people read an answer without clicking. To measure mentions, write down the 20 to 50 questions your buyers actually ask, ask each one several times a week (answers vary from run to run), and record which businesses the answer names and which pages it cites. Read the result as a rate, such as named in 2 of 3 runs, not as a single yes or no.
For Microsoft Copilot, Bing Webmaster Tools has an AI Performance report that shows how often your pages are cited. For Google's AI features, Search Console includes them in the Performance report's Web search type.
What a healthy picture looks like
Put the three numbers side by side each month. Crawler fetches tell you AI search can read you. AI Assistant sessions tell you it is sending people. Mention rates tell you how often you are the answer. A site with plenty of fetches and no visits usually has an answer problem: the pages are readable but not the best answer to the question. A site with visits but few fetches is often blocking a search crawler by accident.
If you would rather have this measured and fixed for you, that is what cite-met's plans do. Otherwise, the steps above cost nothing but an afternoon.
Scanner checks this post covers
- CM-104 Search and AI search crawlers allowed: Search crawlers allowed means your robots.txt lets Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot and PerplexityBot fetch your pages.
