Short answer
Most AI crawlers and fetchers never run the JavaScript tag that tools like GA4 rely on, and Google Analytics also filters out known bots automatically. To see AI agent traffic, look where every request is recorded: your server or hosting logs, or your CDN's bot reports such as Cloudflare AI Crawl Control. Use GA4's AI Assistant channel for the people who click through from AI answers.
Key takeaways
- JavaScript analytics rarely record AI crawlers or fetchers at all.
- Server logs and CDN reports see every request, including bots.
- GA4's AI Assistant channel counts people arriving from AI apps.
Why doesn't GA4 show AI agent visits?
Because GA4 only counts visits that run its JavaScript tag, and most AI bots never run it. Google Analytics also removes known bots from your reports on purpose. So AI agent traffic is filtered out twice before you ever see a chart.
The first filter is JavaScript. Tools like GA4 work by loading a small script in the visitor's browser, which then reports the page view. A bot that downloads your HTML but does not run scripts never sends that report. In a December 2024 study using data from MERJ and its own network, Vercel found that "none of the major AI crawlers currently render JavaScript," naming OpenAI's crawlers, Anthropic's ClaudeBot and PerplexityBot among them. The exception it noted was Gemini, which uses Googlebot's infrastructure for rendering.
The second filter is deliberate. Google says that in Google Analytics "traffic from known bots and spiders is automatically excluded." It identifies them with its own research and the Interactive Advertising Bureau's list of known bots. Google also says you cannot turn this off or see how much was removed.
The result: a store can be read thousands of times a month by AI crawlers and fetchers and see none of it in GA4.
What kinds of AI visits are you missing?
There are three kinds, and analytics tools miss them in different ways.
Crawlers collect pages in bulk for training or for an AI search index. Examples are GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot. They fetch HTML and do not run your tag.
User-triggered fetchers load one page because a person asked an AI a question. OpenAI says ChatGPT-User handles actions started by users in ChatGPT. Perplexity says Perplexity-User fetches a page when a user asks a question, so it can answer and link to the source. Anthropic says Claude-User may access websites when people ask Claude questions. These are often the most valuable AI visits, because a real person is behind each one. See what a user-triggered fetcher is.
Browser agents drive a real browser. OpenAI describes the cloud browser behind ChatGPT agent as "a browser running on a computer in the cloud" that can read pages, click buttons and interact with your site. Because it is a browser, it may run your analytics tag. If it does not identify itself in a way your analytics tool recognises, it can look like an ordinary visitor. OpenAI says its signed requests carry a Signature-Agent header set to "https://chatgpt.com", but GA4 does not report request headers.
So JavaScript analytics tend to miss the first two kinds entirely and blend the third in with people.
How do you find AI agents in your server logs?
Search your server or hosting access logs for the bot names AI companies publish. Every request reaches the server, whether or not it runs JavaScript, and most access logs record each request's user agent string.
Useful tokens to search for include:
| Token | Company | Kind |
|---|---|---|
| GPTBot | OpenAI | Training crawler |
| OAI-SearchBot | OpenAI | Search crawler |
| ChatGPT-User | OpenAI | User-triggered fetcher |
| ClaudeBot | Anthropic | Training crawler |
| Claude-SearchBot | Anthropic | Search crawler |
| Claude-User | Anthropic | User-triggered fetcher |
| PerplexityBot | Perplexity | Search crawler |
| Perplexity-User | Perplexity | User-triggered fetcher |
OpenAI publishes the full strings. For example, ChatGPT-User visits as:
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot
Two cautions. First, a user agent is a claim, not proof. Anyone can copy one. To check a request, compare its IP address with the lists the companies publish, such as https://openai.com/chatgpt-user.json and https://www.perplexity.ai/perplexity-user.json. Our IP range and reverse DNS reference covers the rest. Second, many shared hosts only keep logs for a short time, so export them if you want to compare months.
What does Cloudflare AI Crawl Control show?
If your site is behind Cloudflare, AI Crawl Control gives you a ready-made report of AI crawler activity, and Cloudflare says it is "Available on all plans." It sees requests at Cloudflare's edge, before they reach your server, so JavaScript does not matter.
According to Cloudflare's docs, the dashboard has three tabs:
- Overview shows total requests, allowed requests and unsuccessful requests. Total referrals are shown on paid plans only.
- Crawlers lists each AI crawler with its request count and whether it is currently allowed or blocked.
- Metrics charts activity over time, grouped by crawler, and includes a Most popular paths table showing which pages AI crawlers request most.
Cloudflare's docs note that the top referrers view "is available for customers on a paid plan." The same dashboard is where you allow or block each crawler. Our guide to allowing good AI agents through Cloudflare walks through those settings.
Other CDNs and hosts have their own bot reports. Check your provider's documentation for what it records.
How do you track people who click through from ChatGPT and other AI apps?
Use GA4's AI Assistant channel. This is the traffic your analytics can see, because a person in a normal browser runs your tag.
Google describes the channel as the one "by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok." It notes that it "excludes Google's AI Overviews and AI Mode." Visits from those appear with your other Google traffic.
A visit lands in the AI Assistant channel in one of two ways, according to Google:
- The referrer matches Google's list of AI assistants. GA4 then sets the medium to
ai-assistant. - The link itself carries a medium of exactly
ai-assistant.
The second rule is useful for links you control. If you publish links that AI tools read, such as the URLs in an llms.txt file, you can tag them with standard campaign parameters. Google lists utm_source as the referrer and utm_medium as the marketing medium. For example:
https://www.example.com/products/oak-desk?utm_source=llms-txt&utm_medium=ai-assistant
Keep the tags short and consistent, and never put customer data in them.
Remember what this channel measures. It counts people who clicked a link in an AI answer. It does not count the AI fetch that produced the answer, which happened earlier and left no trace in GA4.
How should you put the picture together?
Read the three sources side by side. Each one answers a different question.
- Server logs or CDN reports: which AI bots and agents read your site, how often and which pages.
- GA4's AI Assistant channel: how many people then arrived from an AI app, and what they did.
- Orders and leads: whether either of those turned into revenue.
A rise in user-triggered fetches for a product page, followed by more AI Assistant sessions to the same page, is a fair sign that AI answers are recommending it. A rise in crawler requests with no matching visits may mean you are being indexed or used for training without much traffic in return. That is a reason to review your robots.txt for AI agents, not proof of anything on its own.
How Sill helps
Sill shows you which AI agents and bots visit your site, the pages they read, and what agents ask for and buy. On Cloudflare Worker and WordPress plugin installs, Sill records agent requests at the server, including the ones that never run JavaScript. In the dashboard, traffic is split into AI agents, other bots and people, so known AI visits are not buried in your human numbers. You can start with a free readiness scan at /scan, no account needed.
Questions people ask
Can I turn off GA4's bot filtering to see AI bots?
No. Google says you cannot disable known bot traffic exclusion in Google Analytics, or see how much known bot traffic was excluded.
Does GA4 have a channel for ChatGPT traffic?
Yes. GA4's default channel group includes an AI Assistant channel for visitors arriving from sources like ChatGPT, Gemini, Deepseek, Copilot or Grok. Google says it excludes Google's AI Overviews and AI Mode.
Is Cloudflare AI Crawl Control free?
Cloudflare says AI Crawl Control is available on all plans. Some parts, such as referral counts and the top referrers list, are for paid plans only.
Why does Google-Extended never show up in my logs?
Google-Extended is a robots.txt control token, not a separate crawler, so it never appears as a visitor. Google's existing crawlers do the fetching.
Sources
- Google Analytics Help: [GA4] Default channel group · checked 17 Nov 2026
- Google Analytics Help: [GA4] Known bot-traffic exclusion · checked 17 Nov 2026
- Google Analytics Help: URL builders: Collect campaign data with custom URLs · checked 17 Nov 2026
- Vercel: The rise of the AI crawler · checked 17 Nov 2026
- Cloudflare docs: AI Crawl Control · checked 17 Nov 2026
- Cloudflare docs: Analyze AI traffic · checked 17 Nov 2026
- OpenAI: Overview of OpenAI crawlers · checked 17 Nov 2026
- OpenAI: Help ensure cloud browser access to your site · checked 17 Nov 2026
- Perplexity: Perplexity crawlers · checked 17 Nov 2026
- Google: Google's common crawlers · checked 17 Nov 2026
- Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler? · checked 17 Nov 2026