What is my user agent?
Below is the user agent your browser sends to every website, and what it says. Paste any other user agent to find out whether it is a crawler, an automated browser or a person.
The user agent you check is sent to Privatrak’s API only to classify it, and is not stored or logged.
What a user agent is
A user agent is a line of text your browser sends with every request to a website. It names the browser, its version and the operating system, so the site can allow for differences between them. Crawlers and other programs send one too, and that is how they tell a site who they are.
This is the user agent of Chrome on Windows, part by part:
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/141.0.0.0 Safari/537.36
| Part | What it says |
|---|---|
| Mozilla/5.0 | A historical token. Nearly every browser starts with it, so it no longer says anything. |
| (Windows NT 10.0; Win64; x64) | The operating system and processor, here Windows on a 64-bit processor. Windows 11 sends this too. |
| AppleWebKit/537.36 (KHTML, like Gecko) | The engine that draws the page. Chrome’s engine grew out of WebKit and still names it, with a version that no longer changes. |
| Chrome/141.0.0.0 | The browser and its main version. Chrome sends zeros for the rest of the version number. |
| Safari/537.36 | Another borrowed name, so that sites written for Safari work in Chrome as well. |
Mozilla/5.0
- What it says
- A historical token. Nearly every browser starts with it, so it no longer says anything.
(Windows NT 10.0; Win64; x64)
- What it says
- The operating system and processor, here Windows on a 64-bit processor. Windows 11 sends this too.
AppleWebKit/537.36 (KHTML, like Gecko)
- What it says
- The engine that draws the page. Chrome’s engine grew out of WebKit and still names it, with a version that no longer changes.
Chrome/141.0.0.0
- What it says
- The browser and its main version. Chrome sends zeros for the rest of the version number.
Safari/537.36
- What it says
- Another borrowed name, so that sites written for Safari work in Chrome as well.
Much of it is history. Browsers copied each other’s names so that sites built for one would also work in the others, and the habit stuck.
How crawlers name themselves
A crawler is a program that reads web pages on its own, for a search engine, an AI service or a tool. Most crawlers put their name in the user agent, often with a link to a page about them, such as compatible; GPTBot/1.3; +https://openai.com/gptbot. The checker compares the user agent with Privatrak’s list of crawler names, which builds on the open crawler-user-agents list.
AI companies run three kinds of crawler, each with its own job:
- Training. GPTBot and ClaudeBot collect pages to train AI models.
- Search. OAI-SearchBot, Claude-SearchBot and PerplexityBot read pages for the search index an AI service answers from.
- Answers for one person. ChatGPT-User and Claude-User read a page when someone asks ChatGPT or Claude a question, while that person waits for the answer.
Which company runs which crawler, and what blocking each one does, is in ClaudeBot, GPTBot, ChatGPT-User: which AI crawler does what.
A user agent is a claim, not proof
Anyone can send any user agent. A scanner looking for weak spots on a site can call itself ClaudeBot, and anyone can set their browser to say it is Googlebot. The checker reads only the text, so it tells you what a request says it is.
To know whether a request really came from a crawler, compare the address it came from with the IP ranges the company publishes. How to check the IP address is explained in our article on AI crawlers.
For the same reason, Privatrak names the AI service a crawler reads for only when that company’s own documentation names the user agent. A user agent that claims to read for an AI service without that is shown as unconfirmed.
How Privatrak counts each kind
Privatrak sorts every visit by its user agent with the same rules as this checker, before anything is counted:
- Crawlers are not counted as visitors and do not use your monthly event limit. Privatrak lists which pages they read on the AI services page under All crawlers.
- Automated visits come from a browser driven by a program, such as a test tool or an AI agent. They count like any visit and use your event limit. Privatrak marks them, so you can leave them out with the filter Automated visit in Insights and on the Events page.
- Everything else counts as a visitor. That includes tools like curl, because many sites send events to Privatrak from their own servers with them.
When a crawler name could also match a browser a person uses, Privatrak leaves it off the crawler list. A crawler that slips through costs a few events. A person left out is data you never get back. More in How visitors are counted.
Many AI crawlers don’t run JavaScript, so a tracker on the page never sees them. Your server can tell Privatrak about their requests instead, as Report server requests shows.
User agent checker questions
What is my user agent?
It is the line of text your browser sends with every request to a website. It names your browser, its version and your operating system. The box at the top of this page shows yours, read from your browser, and below it what it says.
How do I change my user agent?
Browsers let developers change it for testing. In Chrome and Edge, open the developer tools, then More tools and Network conditions, and choose another user agent there. It applies while the developer tools are open. In Safari, it is in the Develop menu under User Agent. Browser extensions can change it too. A changed user agent changes what sites see, not how your browser works.
Why does Chrome’s user agent look frozen?
Chrome and other browsers built on Chromium shortened their user agent to make it harder to tell people apart. They send only the main version number followed by zeros, Windows NT 10.0 on both Windows 10 and Windows 11, Mac OS X 10_15_7 on every Mac and Android 10 with K in place of the phone model. Sites that need the exact details can ask for them through User-Agent Client Hints.
Can I stop GPTBot or ClaudeBot from crawling my site?
Yes, with your robots.txt. Add the line User-agent: GPTBot followed by Disallow: / and do the same for ClaudeBot. OpenAI and Anthropic say their crawlers follow robots.txt. Blocking a training crawler does not take your site out of ChatGPT’s or Claude’s search and answers, because those use other crawlers. A scanner that only borrows the name ignores robots.txt.
Does my user agent identify me personally?
Not on its own. Many people send exactly the same user agent, and Chrome shortened its own to make that more likely. Together with other details, such as your IP address, it can help tell visitors apart. Privatrak uses it that way only for a moment, mixed with a key that is deleted every night, and never stores it.
Why is curl not counted as a crawler?
Many sites send events to Privatrak from their own servers with tools like curl or a programming library. Treating those as crawlers would drop real events, so Privatrak counts them like any visit.