Skip to main content

User-Agent Parser

Paste a User-Agent string to identify the browser, rendering engine, operating system and device, including modern AI crawlers such as GPTBot and ClaudeBot.

User-Agent Parser
BrowserChromev157.0.0.0
EngineBlinkv157.0.0.0
OSWindows10 / 11
Device🖥️ Desktop
CPUx86_64 (64-bit)
Frozen values
  • Chrome's User-Agent reduction fixes everything after the major version at 0. Only the major number is real.
  • Windows 11 also reports "Windows NT 10.0", so the string cannot tell Windows 10 from 11.

Break a User-Agent string into the browser, rendering engine, operating system and device it claims to be. Your own browser's string is already filled in when the page opens. AI crawlers are identified specifically, which matters when you are deciding what to allow in robots.txt.

User-Agent Parser: Features

Why User-Agent strings are so strange

Every modern browser claims to be Mozilla, and most claim to be several other browsers as well. This is historical: sites once served different content based on the User-Agent, so each new browser added the identifiers of its predecessors to avoid being served a degraded page. The result is a string that accumulates lies for compatibility, which is why parsing it requires knowing the order in which to look for the markers rather than simply reading it.

Identifying AI crawlers

The parser recognises the AI crawlers by name: GPTBot, OAI-SearchBot and ChatGPT-User from OpenAI; ClaudeBot, Claude-SearchBot and Claude-User from Anthropic; PerplexityBot and Perplexity-User; the Meta, Amazon and Mistral AI crawlers; Google-Agent; and CCBot from Common Crawl. It also says whether each one collects training data, indexes for AI search or fetches on a user's request. Google-Extended and Applebot-Extended never appear in a User-Agent, because they are robots.txt tokens only. Whether to allow each crawler is a decision every site makes differently; the User-Agent List gives each one's string and token.

Do not use this for feature detection

Parsing the User-Agent to decide what your code can do is a long-standing mistake. The string can be changed by the user, is routinely spoofed, and tells you what the browser claims rather than what it supports. Feature detection is the correct approach: test for the capability directly. User-Agent parsing is appropriate for analytics, for logging, and for identifying bots, where an approximate answer is genuinely useful.

The string is in decline

Browsers have been reducing the information in the User-Agent for privacy reasons, because the combination of values is a strong fingerprinting signal. Chrome reports only its major version and fixes Android at "Android 10; K"; Safari and Chrome on macOS report 10_15_7 and Firefox 10.15 whatever the real version; Windows 11 still says Windows NT 10.0; and since iOS 26, Safari reports the iOS version as a fixed 18.x value. The parser marks these frozen values with a note below the result. The User-Agent List shows what current browsers actually send.

Parsed locally, including your own User-Agent

The parsing runs locally and the string is never transmitted, stored or logged. Your own User-Agent, filled in when the page opens, is read from your browser directly, and the Copy button puts it on the clipboard if you need to send it to a support desk. To examine the rest of the headers that came with a request, the HTTP Header Analyzer explains each one and flags weak security settings.

User-Agent Parser: FAQ

Why does every browser claim to be Mozilla?

A compatibility legacy from the 1990s. Sites served better content to Netscape, so other browsers began identifying as Mozilla to receive it, and each subsequent browser inherited the convention. The token is now meaningless and simply appears in every string.

Which AI crawlers does it recognise?

GPTBot, OAI-SearchBot and ChatGPT-User from OpenAI; ClaudeBot, Claude-SearchBot and Claude-User from Anthropic; PerplexityBot and Perplexity-User; the AI crawlers of Meta, Amazon, Mistral AI and DuckDuckGo; Google-Agent; and CCBot from Common Crawl. Google-Extended and Applebot-Extended are robots.txt tokens with no User-Agent of their own, so they never appear in server logs.

How do I block an AI crawler?

Add a disallow rule for its token in robots.txt. Well-behaved crawlers, including the major AI ones, honour it. A crawler that ignores robots.txt needs blocking at the server or CDN level by User-Agent or by address range, though both are straightforward to evade.

What are Client Hints?

A replacement mechanism where the browser sends detail such as platform and version only when the server asks for it. They exist because the User-Agent string is a strong fingerprinting signal. The practical effect is that the string carries progressively less information, and precise detection increasingly requires Client Hints.

Why is my browser version wrong or missing?

Because browsers have frozen or reduced parts of the string deliberately. Chrome reports only its major version, Android as "Android 10; K" and macOS as 10_15_7, and Safari on iOS 26 and later reports an 18.x iOS version. This is intentional privacy work rather than a parsing failure. The parser flags these values, and for Safari on iOS it reads the real iOS version from the Version token instead.

Can I tell mobile from desktop reliably?

Approximately. The Mobile token is a reasonable signal and tablets are genuinely ambiguous, with iPads long reporting as desktop by default. For layout, use CSS media queries and feature detection rather than the User-Agent; for analytics, the approximate answer is fine.

How to Use User-Agent Parser

  1. Paste a User-Agent, or use your own

    Your own browser's User-Agent is filled in when the page opens. Paste a string from your server logs to see how that one parses instead.

  2. Read the breakdown

    The browser, engine, operating system and device are shown separately, along with a bot identification when the string belongs to a known crawler.

  3. Use it for the right purpose

    Logging, analytics and bot identification are good uses. Deciding what your code can do is not: feature-detect instead, since the string can be changed and increasingly omits detail.

Category Network