An ad verification proxy is a proxy, usually residential, that routes ad checks through IP addresses in the locations a campaign targets. It lets the checker see the ad a real user in that market, on that kind of device, would see.
Ad Verification Proxies: How Ad Teams Check Ads by Country, City, and Device
An ad verification proxy routes your check through an IP address in the place you want to test, so you see the ad that a real user there would see. Ads are targeted by country, city, and device, and some bad actors show a clean version to anyone who looks like a reviewer. Checking from your office, or from a cloud server, shows you neither the targeted ad nor the version served to real users. A residential proxy in Chicago on a mobile device shows you what Chicago mobile users get.
Below: why location changes what an ad check sees, what ad teams verify, how proxy types compare for the job, and a five-step workflow using residential proxies and a rendering API.
Key Takeaways
- Ads change with the viewer. A check from the wrong location or device type can't see the ad a campaign actually serves.
- Cloaking hides bad ads from reviewers. A Google-led study found 4.9% of ads for high-risk search terms cloaked against Google's crawler, and 2026 reporting says malicious creatives increasingly activate only under certain geographic or device conditions.
- Residential IPs give you the user's view. Routing checks through real consumer devices in the target location removes the "reviewer" signal of a data center IP.
- Massive targets by country, subdivision, city, ZIP code, and device type, and the Web Render API can render the page or the search results page from that location.
- Verification is a repeat job. Bad actors rotate domains and creatives, so checks run on a schedule, not once.
What is ad verification?
Ad verification is checking that an ad ran where, when, and how it was bought, and that it's what it claims to be. Brands, agencies, ad networks, and publishers each do some version of it. The common checks:
- Geo compliance: did the ad show in the countries and cities the campaign targeted, and nowhere else?
- Placement and context: did it appear on the agreed sites and next to brand-safe content?
- Creative and landing page: is the ad the approved creative, and does the click lead where it should, without redirects to something else?
- Malvertising: is an ad running on a site delivering a scam or malicious redirect to some visitors?
- Competitive and search ads: which ads appear for a keyword in a given market?
Every one of these depends on seeing the ad from the right place. That's the job of the proxy.
Why do ads look different by location and device?
Because the ad is chosen for each viewer at the moment the page loads. Ad platforms decide what to serve from the request: the IP's location, the device and browser, the time, and more. A campaign targeted at mobile users in Texas simply won't serve to a desktop browser in London. To check it, the request has to come from a mobile device context in Texas.
The harder problem is that some advertisers don't want to be checked. Cloaking means showing one version of a page or ad to reviewers and another to real users. In a 2016 study, Google researchers built a system to detect "split-view content returned to two or more distinct browsing profiles" and applied it to more than 135,000 search and ad URLs for high-risk terms. It found that 4.9% of ads and 11.7% of the top 100 search results cloaked against Google's crawler.
The researchers concluded that security crawlers need, at minimum, ways to get past mobile, reverse DNS, and IP-based cloaking (Invernizzi et al., IEEE S&P 2016).
The pattern hasn't gone away. Reporting on The Media Trust's 2026 Digital Advertising Intelligence Report says malicious creatives are increasingly designed to "activate only under particular geographic, device, or behavioral conditions" (Cybersecurity News, August 2026). And the volume is large: Google says it blocked or removed more than 8.3 billion ads and suspended 24.9 million advertiser accounts in 2025, including 602 million ads tied to scams (Google 2025 Ads Safety Report).
A check from a data center IP is the easiest kind of reviewer to spot. A residential IP in the right city, on the right device type, sees what users see.
What should an ad verification proxy offer?
Three kinds of proxy are commonly used, and they differ on the things that matter for this job:
Beyond the proxy type, look for:
- Fine-grained location. Country isn't enough for local campaigns. You want subdivision (state or region), city, and ideally ZIP code targeting.
- Device type. Mobile ad inventory differs from desktop, and in-app and mobile web differ again, so being able to route through mobile devices matters.
- Real consumer IPs, sourced with consent. Ask how the provider's devices joined the network. Every device on Massive's network opted in through the Massive SDK (how Massive's opt-in network works).
- Session control. Following an ad from impression to landing page should stay on one IP, so you need sticky sessions with a set duration.
- Rendering. Many ads only appear after JavaScript runs, so you need a real browser or a rendering API, not just raw HTML.
A five-step ad verification workflow
1. Build the test matrix
List the markets, device types, and placements each campaign targets, plus a few it shouldn't reach. For a hypothetical US retail campaign targeted at mobile users in Illinois and Texas, a starting matrix might look like this:
Each row is one routed check, repeated on a schedule. (For plans and pricing for this use case, see Massive's proxies for ad verification page.)
2. Route each check from the target location and device
With Massive's residential proxies, location and device type go in the proxy username (more on city, ZIP, and ASN targeting). This routes a request through a mobile device in Chicago, on one IP for 30 minutes:
Massive supports country, subdivision, city, and zipcode targeting, and type values of mobile, common (any non-mobile device), and tv (geotargeting, device-type targeting). Sticky sessions last 15 minutes by default and up to 24 hours with sessionttl (sticky sessions).
3. Render the page and capture the ad
If you'd rather not run browsers, Massive's Web Render API browser endpoint renders the page from the location you choose, with optional device emulation, and returns the rendered HTML. For search ads, the search endpoint takes the same country, subdivision, and city parameters and can wait for sponsored results to load with awaiting=ads (Search docs). Keep the creative, the ad's destination URL, and the time of each check.
Rendering matters more than it looks. On November 15, 2026, we fetched eight US news, sports, and weather homepages twice from the same network, once with a desktop browser user agent and once with an iPhone user agent, and listed the known ad-tech services each HTML response referenced:
On all seven pages that loaded, both devices got the same ad stack. So on these sites, the choice of which ad to show happens after the page arrives, when that code runs and calls the ad servers. A check that reads only the HTML can't see that choice. One that renders the page from the target device and location can.
4. Follow the click
Follow the ad's landing URL from the same session and record every redirect. Cloaking can sit in the redirect chain rather than in the creative itself, which is one reason the Google researchers compared what different browsing profiles were sent. Compare the final landing page with the one the advertiser declared.
5. Compare against the campaign, and repeat
Flag ads that served out of geo, on the wrong device type, next to blocked content, or with a landing page that doesn't match. Then schedule the checks again. Bad actors change creatives and domains quickly, so a single pass misses what changes after it.
Keep it inside the rules
Ad verification exists to protect advertisers and users, and the same proxies must not be used to generate fake impressions or clicks. Massive's network blocks a published set of destination categories for every account and only opens the standard web ports 80 and 443 by default (usage restrictions). We explain how that works in how domain blocking works on Massive's network. The same vantage-point approach also works for monitoring ads in AI assistants, which we covered in how to monitor ChatGPT ads.
In short, an ad verification proxy earns its keep by making each check look like the viewer the campaign was bought for: the right city, the right kind of device, a real consumer connection. Build the matrix, render from each row, follow the click, and repeat.
Sources
- Invernizzi, Thomas, Kapravelos, Comanescu, Picod, and Bursztein, Cloak of Visibility: Detecting When Machines Browse a Different Web, IEEE Symposium on Security and Privacy, 2016
- Cybersecurity News, Malvertising Is Moving From Deceptive Content to Weaponized Infrastructure, August 29, 2026, reporting on The Media Trust's 2026 Digital Advertising Intelligence Report (released March 4, 2026)
- Massive, desktop vs mobile fetch of eight US news, sports, and weather homepages, November 15, 2026 (first-party check; script and saved HTML available on request)
- Google, 2025 Ads Safety Report, April 16, 2026
- Massive documentation: geotargeting, device-type targeting, sticky sessions, Search, usage restrictions
Frequently Asked Questions
Data center IPs don't match where real users browse, and they're easy for cloaking systems to recognize as reviewers. Residential IPs come from consumer connections in the target location, so the ad served is the one users get.
Ad verification can catch much of it, though not all. Comparing what different browsing profiles are sent is how cloaking is found; that's the method of the 2016 Google study. The researchers concluded that a crawler needs, at minimum, ways past mobile, reverse DNS, and IP-based cloaking, which is what checks from residential, mobile, and in-market IPs provide.
Continuously, or at least on a fixed schedule for each live campaign. In practice, that means re-running every row of the campaign's test matrix (each target city, device type, and control) at set intervals, and again whenever creatives or targeting change. Malicious actors change creatives and domains frequently, so a one-time check is not enough.
