• Возможности
  • Тарифы
  • Вопросы
  • Блог
ВойтиНачать бесплатно
0%
Binetro

Защита

  • Защита от ботов
  • Защита от скликивания

Аналитика

  • Аналитика трафика
  • Отчёты

Подключение

  • Тильда
  • WordPress
  • Любой сайт

Начало работы

  • Как подключить
  • Тарифы
  • Блог

Документы

  • Политика конфиденциальности
  • Договор публичной оферты
  • Согласие на обработку данных
  • Политика использования cookie
  1. Blog
  2. Bots and technology

What a proxy is, and what a residential proxy is not in these sources

Published October 25, 2026Binetro editorial teamBots and technology6 min read—
Contents
  1. A forward proxy and a tunnel
  2. One external address for many
  3. Invalid proxy traffic in ad measurement

Contents

  1. A forward proxy and a tunnel
  2. One external address for many
  3. Invalid proxy traffic in ad measurement

A high CTR isn't success yet

Binetro separates real visits from bot clicks and protects your ad campaign.

Protect your ads

Share

A forward proxy acts on behalf of a client or a group of clients and can hide their address. A reverse proxy hides the server and is used for load balancing, cache, and compression. That is how MDN splits proxies. The phrase “residential proxy” is not on that page. It is also absent from the Carrier-Grade NAT discussion and from the Google Ads help on location and IP. How to look at traffic by address and provider is covered in how to check traffic by IP and provider. Below is what these sources say instead of a definition they do not have.

A data-center building with rows of housings
Under CGN, geolocation sees the translation server, not the end device. · Photo: CC BY 2.0, Tony Webster from Minneapolis, Minnesota, United States

A forward proxy and a tunnel

The CONNECT method opens a tunnel: that is how a client behind an HTTP proxy reaches a site over TLS. Not every proxy accepts CONNECT. Some cut it before port 443. The Forwarded and X-Forwarded-For headers can carry the client’s original address, so a proxy does not always hide it.

Hiding the address is a possibility, not a promise. If the header travels further, the site sees the original address next to the proxy’s. If the proxy did not accept CONNECT, there is no TLS tunnel to the site. Neither case is called a sign of being “residential” on that page. There is no class by that name there.

One external address for many

RFC 6598 (April 2012) assigns providers a shared address space for Carrier-Grade NAT: the block 100.64.0.0/10. It is separate from the private addresses of RFC 1918 because it is meant for the provider’s network. Packets with these addresses must not be forwarded past the provider’s border.

Under CGN, geolocation systems locate the translation server, not the end device. Some sites limit how many simultaneous logins share one external public IPv4. Several subscribers share that external address. In the example with game consoles, two subscribers with one external address cannot connect to each other. The RFC gives no count of subscribers per address. The example is a class of application breakage, not a measurement of an audience. The words “residential proxy” and “thousands of people” are not there.

The Google Ads glossary calls an IP address a number a provider assigns to a device on the internet. Sometimes a provider assigns the same address to a large number of computers, and different computers share it. The page does not attach a figure to “a large number”. The address is often used to judge where the computer reaches the network, and that signal takes part in geographic targeting. The same page says an IP exclusion may stop an ad from showing to a specific address. That is not permission to exclude shared addresses.

Google Ads usually derives a physical place from the IP address the provider assigns to the computer. If the device is on a Wi-Fi network, they may take that network’s address. If a mobile device exits through the carrier’s proxy, they may take the carrier’s address. When location settings are on, GPS, Wi-Fi, Bluetooth, and a cell-tower database are added. The help page qualifies the accuracy of each source separately. Interest in a place may be derived from search wording, past queries, and past physical places. Targeting is a best effort given settings, devices, and behavior, and one hundred percent accuracy is not guaranteed. The page does not use the word VPN and does not call the carrier’s proxy a residential proxy.

Invalid proxy traffic in ad measurement

The MRC addendum places invalid proxy traffic among SIVT examples: situations that are harder to notice and that need advanced analysis, a check of several signs, or a noticeable human part. Beside it in the same examples are automation from a dedicated or captured device, motivated human activity, forced and accidental clicks from a swapped interface, fake measurement events, and bots that disguise themselves. A footnote gives a click farm as organized non-robotic activity through many users or machines. IVT covers GIVT and SIVT together. Some SIVT can move into GIVT over time if standard lists appear.

That is a classification for measuring advertising, not a description of a proxy market and not a procedure. The document asks not to mix disclosed reward programs that do not aim to substitute measurement into this category. “Invalid proxy traffic” here is the name of an example in a standard, not a synonym for every proxy and not a definition of a residential one.

SourceWhat it definesWhat it does not say
MDNa forward proxy on the client’s behalf, a reverse proxy on the server’sa residential proxy
RFC 6598a shared external IPv4 for CGN subscribers; geolocation sees the serverhow many people per address
Google Adsplace from IP, Wi-Fi, a carrier proxy, GPS, and past queriesthat a carrier proxy is residential
MRCinvalid proxy traffic as an example of hard IVThow the service is built

A high CTR isn't success yet

Binetro separates real visits from bot clicks and protects your ad campaign.

Protect your ads

Related

  • How modern bots work: from scripts to AI agents

    How a Yandex Metrica robot differs from SIVT, a windowless browser, a search crawler, and a visit triggered by a chat question. A way to keep those cases apart.

    11 min read
  • Why a captcha does not stop bots

    A captcha in Yandex Direct and in the OWASP catalog slows automation, but the test can be solved. How that differs from a browser proof-of-work task.

    6 min read
  • How click bots that imitate behavior work

    In public descriptions a click bot is a program that imitates an ordinary browse for a click or an impression. How that differs from a windowless browser, a botnet, and repeated human clicks.

    6 min read
  • What a bot farm is

    In the MRC standard a click farm is organized non-robotic activity through many people or machines. How SIVT differs from routine GIVT, and why a solved captcha does not prove a person.

    5 min read
All articles