User agents in web scraping play a major role in preventing scraper blocking. In this article, we'll take a look at what user agents are, why they are important in web scraping and how to rotate user agents for web scraping in python.
What is User-Agent Header?
The User-Agent header is one of many standard request headers. It identifies the request sender's device which includes information about the device type, operating system, browser name and version.
Historically, the User-Agent was intended to help websites optimize served content for different devices. However, contemporary responsive websites mostly use this datapoint for tracking.
The value of the user agent header is called user agent string. It follows a straightforward pattern that provides various details about the connecting user:
Mozilla/5.0: A historical artifact (always there).
<system-information>: Represents the operating system, its version and the CPU architecture.
<platform>: Represent the rendering engine used and its version.
<browser-info>: Provides information about the used web browser and it's version.
This pattern creates user agent strings that look like this:
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0
To parse this User-Agent string and extract information, we can use Python ua_parser package which can extract user agent details from a given user agent string like so:
from ua_parser import user_agent_parser
user_agent_string = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0"
parsed_string = user_agent_parser.Parse(user_agent_string)
print(parsed_string)
{
"device": {
"brand": "Apple",
"family": "Mac",
"model": "Mac",
},
"os": {
"family": "Mac OS X",
"major": "10",
"minor": "15",
"patch": None,
"patch_minor": None,
},
"string": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0",
"user_agent": {
"family": "Firefox",
"major": "113",
"minor": "0",
"patch": None,
},
}
Using ua_parser we can clearly see that this user agent represents a Firefox browser with version 113 running on a Mac device with version 10.15 and uses the Gecko rendering engine.
Now that we understand User-Agent header's purpose and structure, let’s see why we should use user agents for web scraping.
Many websites are likely to block requests that don't provide any User-Agent headers. While most throttle or block specific User-Agent headers that appear unusual like Linux operating system or unknown web browser.
To prevent scrapers from being blocked because of user agent it's best to rotate User-Agent headers for each outgoing request. Let's take a look at how to establish an user-agent pool and randomly rotate them in web scraping next.
How to Get User-Agent Strings?
The easiest way to get a User-Agent header is to obtain it from a real web browser using developer tools. Open any web page in browser like Chrome or Firefox and press the (F12) key to open Browser Developer Tools, then select the Network tab and reload the page:
Each request made by the browser includes a set of headers values and one of them is the user agent.
We can use this user agent for small scale web scraping. Though, when scraping at scale we'll need more than one user agent string to prevent scraper identification and blocking.
User-Agent Scraping
To scale up, we need to collect a dataset of user agent strings and rotate them semi-randomly for the best scraping results.
There are many online databases for user agent strings though in this example we'll scrape a list of user agents from useragentlist.net and rotate them using weighted randomization in Python.
useragentlist.net is one of many online user agent string databases
To scrape the user-agent strings, we'll use httpx alongside with BeautifulSoup. We'll also use ua_parse to parse user-agent strings to user agent objects we can rotate in Python. All of these libraries can be installed using pip console command:
$ pip install httpx beautifulsoup4 ua-parser
To scrape user agent strings from useragentlist.net online database we'll use httpx to retrieve the HTML page and bs4 to parse the HTML and extract the user agents:
from bs4 import BeautifulSoup
import httpx
url = "https://www.useragentlist.net/"
request = httpx.get(url)
user_agents = []
soup = BeautifulSoup(request.text, "html.parser")
for user_agent in soup.select("pre.wp-block-code"):
user_agents.append(user_agent.text)
print(user_agents)
Here, we send a request to the target website and wrap the HTML into a BeautifulSoup's soup object. We then search for all user agent strings on the page and return the result. Here is what we got:
Output result
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 6.3; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 10.0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 6.3; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Windows NT 10.0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Windows NT 6.1; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Windows NT 10.0; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Android 12; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0
Mozilla/5.0 (Android 13; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0
Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Android 11; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0
Mozilla/5.0 (Windows NT 6.2; Win64; x64; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Windows NT 6.3; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.13; rv:109.0) Gecko/20100101 Firefox/113.0
Mozilla/5.0 (Android 10; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36
Mozilla/5.0 (Windows NT 6.3; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36
Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36
Mozilla/5.0 (Linux; Android 10; Lenovo TB-8505F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; Infinix X656) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; LM-Q730) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; M2004J19C) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; SM-N960F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 11; A509DL) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 11; moto g pure) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 11; SM-A115F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 11; SM-A207F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 11; SM-A207M) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; FNE-NX9) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; M2101K7AG) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; motorola edge 5G UW (2021)) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; SM-A115F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; SM-A135U) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; SM-M515F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 12; SM-S127DL) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 13; SM-A536E) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 9; INE-LX2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 9; SM-J530F) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; COL-L29) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; CPH1819) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; CPH1931) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; CPH2179) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; ELE-L29) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; HRY-LX1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Mozilla/5.0 (Linux; Android 10; JSN-L21) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Mobile Safari/537.36
Now that we have a list of user agents for scraping. Let’s move on to the user agents rotation.
How to Rotate User-Agent Headers
The idea behind rotating user agents is to change the user agent for every request being made by the scraper. This will prevent user-agent based blocking as the traffic splits between all user agents.
We could just grab a random user agent each time however not all user agent strings are of the same value. For example:
Newer browser versions are more likely to be successful.
Windows and Macos operating systems are more likely to be successful.
More popular web browsers like Chrome and Firefox are more likely to succeed.
The rotation criteria will be defined based on weights, where user agents with higher weights will get higher priority. The weighting criteria is defined based on several factors, including the device, operating system and web browser. For example, user agents for Mac and Windows devices will get higher weights than Linux and Android devices.
Let’s put that into code by creating a user agent container object and a rotator class:
from ua_parser import user_agent_parser
from functools import cached_property
from typing import List
from time import time
import random
class UserAgent:
"""container for a User-Agent"""
def __init__(self, string) -> None:
self.string: str = string
# Parse the User-Agent string
self.parsed_string: str = user_agent_parser.Parse(string)
self.last_used: int = time()
# Get the browser name
@cached_property
def browser(self) -> str:
return self.parsed_string["user_agent"]["family"]
# Get the browser version
@cached_property
def browser_version(self) -> int:
return int(self.parsed_string["user_agent"]["major"])
# Get the operation system
@cached_property
def os(self) -> str:
return self.parsed_string["os"]["family"]
# Return the actual user agent string
def __str__(self) -> str:
return self.string
class Rotator:
"""weighted random user agent rotator"""
def __init__(self, user_agents: List[UserAgent]):
# Add User-Agent strings to the UserAgent container
user_agents = [UserAgent(ua) for ua in user_agents]
self.user_agents = user_agents
# Add weight for each User-Agent
def weigh_user_agent(self, user_agent: UserAgent):
weight = 1_000
# Add higher weight for less used User-Agents
if user_agent.last_used:
_seconds_since_last_use = time() - user_agent.last_used
weight += _seconds_since_last_use
# Add higher weight based on the browser
if user_agent.browser == "Chrome":
weight += 100
if user_agent.browser == "Firefox" or "Edge":
weight += 50
if user_agent.browser == "Chrome Mobile" or "Firefox Mobile":
weight += 0
# Add higher weight for higher browser versions
if user_agent.browser_version:
weight += user_agent.browser_version * 10
# Add higher weight based on the OS type
if user_agent.os == "Windows":
weight += 150
if user_agent.os == "Mac OS X":
weight += 100
if user_agent.os == "Linux" or "Ubuntu":
weight -= 50
if user_agent.os == "Android":
weight -= 100
return weight
def get(self):
# Weigh all User-Agents
user_agent_weights = [
self.weigh_user_agent(user_agent) for user_agent in self.user_agents
]
# Select a random User-Agent
user_agent = random.choices(
self.user_agents,
weights=user_agent_weights,
k=1,
)[0]
# Update the last used time when selecting a User-Agent
user_agent.last_used = time()
return user_agent
Let’s break down the above code.
First, the User-Agent strings are passed to the UserAgent class to extract its properties using the ua_parser module. Then, we create Rotator object which can select a random UserAgent from the pool based on the weights we defined.
Let’s mock a thousand requests with user-agent rotation to see how it works:
from collections import Counter
# Some user agents from the list we created earlier
user_agents = [
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35",
"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35",
"Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35",
"Mozilla/5.0 (Windows NT 6.1; rv:109.0) Gecko/20100101 Firefox/113.0",
"Mozilla/5.0 (Android 12; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0",
"Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0",
]
# The rotator class we created
rotator = Rotator(user_agents)
# Counter to track the most used user agents
counter = Counter()
for i in range(1000):
# Choose a random User-Agent in each iteration
user_agent = rotator.get()
counter[user_agent] += 1
# Show the most used User-Agents
for user_agent, used_count in counter.most_common():
print(f"{user_agent} was used {used_count} times")
We create an instance to the Rotator class to rotate user agents in the list and use the Counter module to track the most used ones. Here is the result we got:
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35 was used 188 times
Mozilla/5.0 (Windows NT 6.1; rv:109.0) Gecko/20100101 Firefox/113.0 was used 184 times
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35 was used 169 times
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101 Firefox/113.0 was used 163 times
Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/113.0.0.0 Safari/537.36 Edg/113.0.1774.35 was used 149 times
Mozilla/5.0 (Android 12; Mobile; rv:109.0) Gecko/113.0 Firefox/113.0 was used 147 times
Cool! We successfully rotated user agents for thousands of requests using weighted randomization which would distribute our scraper traffic through multiple user agent identifiers and prevent blocking.
How to Add User-Agents to Web Scrapers
User-agent string is a header just like any other so to configure it all we have to do is change the User-Agent header in our web scraping client:
from playwright.sync_api import sync_playwright
# Intitialize a playwright instance
with sync_playwright() as playwight:
# Launch a chrome headless browser
browser = playwight.chromium.launch(headless=True)
# Create an independent browser session
context = browser.new_context()
context.set_extra_http_headers(
{
# Add User-Agent header
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36",
}
)
page = context.new_page()
page.goto("https://httpbin.dev/user-agent")
page_content = page.content()
browser.close()
print(page_content)
# {"user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36"}
from scrapfly import ScrapeConfig, ScrapflyClient, ScrapeApiResponse
scrapfly = ScrapflyClient(key="Your ScrapFly API key")
api_response: ScrapeApiResponse = scrapfly.scrape(
scrape_config=ScrapeConfig(
url="https://httpbin.dev/user-agent",
headers={
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36",
},
)
)
page_content = api_response.scrape_result["content"]
print(page_content)
# {"user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36"}
We have seen that we can generate and rotate user agents for web scraping. However, this process is quite complicated and time-consuming. Let’s take a look at an easier solution!
Automatic User-Agents using ScrapFly
User Agents can be quite straight forward when it comes to web scraping but still not enough to scale up some web scraping operations and this is where Scrapfly can lend a hand!
Instead of manually configuring user agents for web scraping ourselves, ScrapFly will automatically manage User-Agent headers and proxies to ensure the highest scraping success rate.
Here's an example of a scraping request with automated user-agent rotation and Scrapfly Python SDK:
from scrapfly import ScrapeConfig, ScrapflyClient, ScrapeApiResponse
scrapfly = ScrapflyClient(key="Your API key")
api_response: ScrapeApiResponse = scrapfly.scrape(
# Send a request to this URL, which returns the used User-Agent header
scrape_config=ScrapeConfig(url="https://httpbin.dev/user-agent")
)
page_content = api_response.scrape_result["content"]
print(page_content)
# {"user-agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36"}
To wrap up this guide on user agents for web scraping, let’s take a look at some frequently asked questions.
What is the User-Agent header?
User-Agent is a request header responsible for representing the request sender's device. It includes information about the device, operating system and the web browser.
How to rotate user agents for web scraping?
The best way to rotate user agent strings in web scraping is to use weighted randomization. As user agent strings vary in quality in web scraping, we can assign weight modifications to each user agent string features like operating system, browser name or version. See this example in python.
What are the most common user agents?
The most common user agents are the ones that represent the most popular web browsers, their versions and operating systems. So, given these 3 details and user-agent pattern we can generate all possible popular user-agent headers!
Why do we need user agents for web scraping?
The User-Agent header is used to add metadata to the request. Websites lookup this metadata to determine if the request sender is a human user or a bot. Therefore, adding user agents to web scrapers helps avoid web scraping blocking.
How can I get a list of user agents for web scraping?
There are multiple online databases for user agents. These online databases can be scraped to generate your own web scraping user-agent pools.
User-Agents for Web Scraping Summary
In this article, we've taken a look at the User-Agent header and how it's used in web scraping. In short, user-agent provides metadata information about the request sender's device and by including appropriate user-agent strings in web scraping requests we can drastically reduce scraper blocking rates.
We've also covered how to rotate user agents for web scraping. We've scraped our own user agent string pool from useragentstrings.net and used weighted randomization to rotate user agents in python. Finally, we've covered how to add these user agents to web scraping clients like httpx, playwright and Scrapfly Python SDK.
