Ad verification cost per check: the bandwidth and proxy math

How much does one ad verification check cost?

On a per-GB residential plan, a check costs whatever data it loads times the plan's rate. Measured on 30 September 2026 from a direct connection, a redirect-chain check that reads whole documents moved 0.8 MB at the median, a full browser render with a screenshot moved 3.8 MB, and a YouTube watch page with eight seconds of playback moved 7.6 MB. On the Residential IPv4 GB-based 1 GB traffic pack ($1.64) that is $1.31, $6.23 and $12.46 per 1,000 checks; on the 5,000 GB traffic pack ($0.60/GB) it is $0.48, $2.28 and $4.56.

Those are medians. The 90th percentile was 5.6 MB, 6.4 MB and 9.7 MB, so budget for the spread, not the middle. This article shows where the numbers come from, how to cut the data a check loads, and when an unlimited plan priced by speed beats paying per gigabyte. The same table sits on the ad verification proxies page and is refreshed with each measurement.

• Redirect-chain check (document requests only): 0.8 MB median, $1.31 per 1,000 on the 1 GB traffic pack rate, $0.48 on the 5,000 GB traffic pack rate.
• Full render with a screenshot: 3.8 MB median, $6.23 and $2.28 per 1,000.
• Video-heavy page (YouTube watch page, 8 s of playback): 7.6 MB median, $12.46 and $4.56 per 1,000.
• Residential IPv4 Unlimited at 10 Mbps moves at most 4.5 GB an hour, so one day ($7.35) caps at about 28,000 full renders at 3.8 MB, country-level targeting only.
• Reading only response headers on a redirect check moves a few kilobytes per hop instead of the whole document.

What a check loads

An ad verification check is a request or a browser session through a proxy exit in the market you are auditing, plus the evidence you keep. What it loads depends on what you need to prove. A redirect-chain check follows an ad click URL or an affiliate link hop by hop and records each status code and host. With a plain HTTP client it fetches only documents: no images, scripts or fonts. Publisher homepages still return large HTML, which is why the median was 0.8 MB and one homepage returned 5.6 MB of HTML alone.

A full render opens the page in a real browser, waits for the network to go quiet, and takes a screenshot. That loads every script, image, ad tag and tracker the page calls, which is the point: the screenshot shows the creative in its slot the way a local visitor sees it. The median across ten publisher homepages was 3.8 MB; nine of fifty loads did not reach the load event within 60 seconds and are excluded, and one publisher served a small bot-wall page instead of its homepage.

A video-heavy page adds media segments. Five YouTube watch pages with autoplay allowed and eight seconds of playback moved 7.6 MB at the median and 12.2 MB at most. A pre-roll ad check that waits for the ad to play will sit at the top of that range.

Measured medians by check type

Method: 30 September 2026, a single direct (non-proxy) connection, headless Chromium 153 driven by Playwright, a fresh browser context for every load so nothing is cached, bytes summed from the HAR transfer sizes. Ten ad-supported publisher homepages (BBC, The Guardian, CNN, Forbes, Weather.com, ESPN, Reuters, Spiegel, Le Monde, The Independent), five loads each for the render and redirect types; five YouTube watch pages, ten loads each, for the video type. The redirect type used a plain HTTP client following up to ten hops. No ad was clicked.

Limits: this pass ran from one direct connection, not from residential exits in each market, so it measures page weight rather than per-country ad payloads. Ad stacks serve different creatives by geography and by network type, and a residential exit in the target country can receive more ad calls than an office connection. The per-country residential pass is the next measurement and the table will be updated with it. Treat these numbers as the floor for a full render and the order of magnitude for the rest.

Cost per 1,000 checks by plan

TrueProxies sells two residential IPv4 billing models. GB-based plans are traffic packs from 1 GB to 5,000 GB, each lasting 30 days from activation: 1 GB for $1.64, down to $0.60/GB on the 5,000 GB traffic pack, with country, city, region and ASN targeting. Unlimited plans bill a speed tier for a duration with no per-GB charge, from $4.08 for one hour at 10 Mbps and $7.35 for a day, with country targeting only. Prices as of 30 September 2026.

The unlimited column is a theoretical ceiling: 10 Mbps moves 4.5 GB an hour and 108 GB a day at 100% utilization, and real throughput depends on the targets, your concurrency and how long each page takes to settle. It also targets by country only, so a check that needs a specific city or ISP belongs on a GB-based plan whatever the volume. The break-even is simple: one day of 10 Mbps unlimited costs the same as 4.5 GB at $1.64/GB or 12.3 GB at $0.60/GB, which is about 1,200 full renders at the small-pack rate or 3,200 at the large-pack rate. Above that, per day, unlimited is cheaper; below it, or for city and ASN checks, a traffic pack is.

Cutting the data per check

• Read headers, not bodies, on redirect checks. A redirect chain is fully described by each hop's status code and Location header. A HEAD request, or a GET that closes the connection after the headers arrive, moves a few kilobytes per hop instead of a multi-megabyte homepage. Only fetch the final document when you need its content.
• Block what the screenshot does not need. For placement checks the ad slots and creatives must load, but fonts, analytics beacons and video preloads usually do not. Blocking by resource type in Playwright or Puppeteer typically removes a third or more of the bytes. Do not block ad-serving hosts, or you measure a page without the thing you are checking.
• Reuse a sticky session for multi-page flows. A landing page followed by a checkout step can share one exit and one browser context, so shared scripts are fetched once. Set the session lifetime to the length of the flow.
• Sample, do not census. Ten checks per market per day, at different hours, show a rotation problem as well as a thousand. Increase the sample only where a difference appears.
• Skip video playback unless the ad is the video. Loading a watch page without autoplay measured about half the bytes of one with eight seconds of playback.

Availability in the markets that ask for ad checks

Search Console shows the ad verification page reached from the United States, Germany, the United Kingdom, Canada, Spain, France, the Netherlands, Japan and Australia. Every one of them is on the published residential IPv4 market snapshot of 12 July 2026, with the pool size and city coverage recorded on its country guide. City, region and ASN targeting is a GB-based plan control written on the proxy username; Unlimited plans target these markets by country.

The cost per check is the same in every market; only the pack size changes the rate. What changes by market is the ad payload, which is why the residential per-country measurement matters and is next on the list.

How to measure your own workload

Run twenty representative checks through the exit type and market you plan to use, record the HAR for each, and take the median and 90th percentile of the transfer sizes. The ad verification evidence kit ships a Playwright script that records a HAR, the redirect chain and a screenshot per check, and the field dictionary tells you which fields to keep. Multiply the 90th percentile by your monthly check count and put the result into the proxy cost calculator to compare traffic packs with unlimited plans for your actual volume.

import { readFileSync } from "node:fs";

const har = JSON.parse(readFileSync("check.har", "utf8"));
let bytes = 0;
for (const entry of har.log.entries) {
  const r = entry.response ?? {};
  const transfer = typeof r._transferSize === "number" && r._transferSize >= 0 ? r._transferSize : null;
  bytes += transfer ?? Math.max(r.bodySize ?? 0, 0) + Math.max(r.headersSize ?? 0, 0);
}
console.log(`${(bytes / 1e6).toFixed(2)} MB over ${har.log.entries.length} requests`);

Record the exit reference and country with every HAR. A check that landed on the wrong exit is a wasted check, not a data point, and the preflight step in the evidence kit method exists to catch it before the page loads.

FAQ

How much bandwidth does ad verification use? Measured on 30 September 2026: about 0.8 MB for a redirect-chain check that reads whole documents, 3.8 MB for a full render with a screenshot and 7.6 MB for a video page, at the median. Header-only redirect checks use a few kilobytes per hop.

Is there a cheap way to verify ads in a specific country? A redirect-chain check through a residential exit in that country costs under $2 per 1,000 on the 1 GB traffic pack rate and under $0.50 per 1,000 on the 5,000 GB traffic pack rate. Full renders cost $2 to $7 per 1,000 depending on the pack. Country-level monitoring at volume fits an unlimited plan from $7.35 a day.

Does a screenshot cost extra bandwidth? No. The screenshot is written locally from the rendered page; the bandwidth is spent loading the page. A full-page screenshot of a long publisher page can, however, trigger lazy-loaded images below the fold, which is part of the 3.8 MB median.

Why measure from a direct connection first? Because page weight is the largest component of the cost and can be measured without spending proxy traffic. The residential per-country pass, which captures how ad payloads differ by market and network type, is the next measurement and will replace the direct medians in the table.

Which plan should I start with? A 1 GB or 3 GB traffic pack covers a few hundred full renders with city and ASN targeting, enough to measure your own median before committing to a larger pack or an unlimited plan. The ad verification proxies page lists the plan choice by check type.