GlossaryWeb ScrapingIntermediate

Rotating Proxy

A rotating proxy is a single gateway endpoint that hands you a different IP address from a shared pool — on every request, or on a timer — so a large job looks like many ordinary visitors instead of one persistent one.

Last updated August 14, 2026

Definition

A rotating proxy is a proxy endpoint that gives you a different exit IP from a shared pool, either on every request or every few minutes. You connect to one address that never changes; the provider quietly swaps the IP behind it.

The value is arithmetic. A site that blocks any visitor making 100 requests an hour will certainly block one address making 10,000. Spread those same 10,000 requests across 5,000 IPs and each one looks like a person who dropped by twice.

How the gateway works

You are not handed a list of IPs to cycle through yourself. The provider gives you a single backconnect gateway — something like gate.provider.com:7000 — and rotates the exit address behind it. Your scraper's configuration never changes; what the target site sees changes constantly.

That design is why rotation is easy to adopt. One host, one port, one set of credentials, and the entire pool is available. It also means the provider is responsible for retiring dead and burned addresses, which is the tedious half of running your own list.

Per request, or per session

Two modes cover nearly every job:

  • Rotate on every request — a new IP for each call. Correct for stateless crawling: product pages, search results, listings, anything where no request depends on the one before it.
  • Sticky sessions — one IP pinned to you for a fixed window, typically 1 to 30 minutes, usually requested by adding a session ID to your proxy username. Required for anything with state: logging in, filling a basket, paginating through a filtered result set.

Choosing wrongly here is the most common rotation mistake. A checkout that begins in Texas and finishes in Poland fails not because the IPs were poor quality but because the journey was impossible.

Rotation is a property of the endpoint, not the pool

"Rotating" describes how addresses are assigned; it says nothing about where they come from. Any pool can rotate, and the pool is what determines trust, speed, and price.

  • Rotating residential — real home ISP addresses, the highest trust and the usual choice for protected targets. Slower and billed by bandwidth.
  • Rotating datacenter — fast and cheap, often sold with unlimited traffic, but easy for anti-bot systems to identify by ASN.
  • Rotating ISP — datacenter speed with residential registration, rotated across a smaller pool.
  • Rotating mobile (4G/5G) — carrier addresses shared by thousands of real handsets through CGNAT, the hardest to block and the most expensive.

What rotation does not fix

A new IP resets one signal. It does not reset the rest of your identity, and modern bot detection weighs several at once.

  • Cookies and local storage carry across a rotation unless you clear them, re-identifying you instantly.
  • Browser fingerprint and TLS fingerprint stay identical whichever IP you exit from.
  • Behaviour — 60 requests a minute in perfect rhythm reads as automation from any address. Rotation should be paired with throttling, not used as a substitute for it.

Rotating an IP while keeping one fingerprint and one cookie jar is actively worse than not rotating, because a single visitor appearing from hundreds of countries is a pattern no real user produces.

What it costs

Rotating residential and mobile pools are billed by gigabyte, so the size of the pages you fetch drives the bill more than the number of requests. Blocking images, fonts, and third-party scripts in your scraper is the single largest saving available. Datacenter pools are usually billed per port or per thread with generous or unlimited traffic, which is why they remain the default for high-volume crawling of soft targets.

Free rotating endpoints exist, and they are shared with everyone else using them — expect addresses already blocked by the sites you care about. Most reputable providers offer a small paid trial instead, which is the honest way to test whether a pool works against your specific targets.

Examples

1

Pointing a scraper at one gateway such as gate.provider.com:7000 and receiving a fresh exit IP on every request

2

Appending a session ID to the proxy username, like user-session-a9f3, to hold a single IP for a ten-minute checkout flow

3

Crawling 50,000 product pages across a residential pool so no individual address ever trips the site's rate limit

4

Switching a Python Requests or Scrapy job to rotation by changing a single line of proxy configuration

5

Using a rotating mobile pool so requests arrive from carrier IPs shared with thousands of genuine phone users

Common Use Cases

High-volume web scraping and competitor price monitoring
Staying under per-IP rate limits without slowing a crawl to a halt
Search engine and marketplace data collection at scale
Ad verification and brand protection from many different exit points
Distributing a large job so traffic resembles many separate users
Reaching sites that block repeated requests from any single address

Frequently Asked Questions

A static proxy gives you one IP that stays the same for as long as you rent it. A rotating proxy gives you a gateway that swaps the exit IP per request or per interval. Static suits long-lived accounts and anything that must be recognised; rotating suits volume.
Only if you use the per-request endpoint. Providers also offer timed rotation and sticky sessions that hold one IP for a set window, and the mode is normally selected by the port you connect to or a parameter in your proxy username.
Whenever requests depend on each other — logging in, completing a checkout, or paging through results behind a filter. Changing IP mid-flow usually invalidates the session and can look like account takeover to the site.
No. Rotation describes how IPs are assigned, not where they come from. Datacenter, ISP, and mobile pools all rotate too. Residential is simply the most common pairing because volume scraping and IP trust tend to be needed together.
No. It removes one reason to block you. Cookies, browser and TLS fingerprints, headers, and request timing all persist across a rotation, and a single fingerprint appearing from hundreds of IPs is itself a strong bot signal.
Yes. Because rotation happens behind one fixed gateway, you set that host, port, and credentials once in your HTTP client or browser launch arguments and never touch it again. No proxy-cycling code is required on your side.
Residential and mobile pools are billed per gigabyte, so page weight matters more than request count — blocking images and scripts cuts the bill sharply. Datacenter rotation is usually sold per port or thread with unlimited or very high traffic caps.