Get started →

Home / Guides / VPN, proxy and data-centre traffic: what publishers should know

● Traffic quality

VPN, proxy and data-centre traffic: what publishers should know

Updated 16 November 2026 · 3 min read · By Arslan Mehdi

When people talk about invalid traffic, three kinds of network come up again and again: data centres, VPNs and proxies. They're not equally risky, and treating them all the same can cost you real readers. Here's how we think about each one.

Why the network matters

Every visit comes from an IP address, and every IP address belongs to a network. Real readers mostly come from home broadband and mobile carriers. Automated browsers mostly run on servers, and servers live in data centres. So the type of network behind a visit is one of the strongest single signals of whether it's a person.

It isn't the only signal, though. Good detection combines the network with device, browser and behaviour checks. We explain the wider categories in invalid traffic explained.

Data-centre traffic

Data-centre traffic comes from cloud and hosting providers. Some of it is legitimate: search engine crawlers, uptime monitors, link previews. Those are known crawlers, and they shouldn't see ads anyway.

The rest is mostly automation: scrapers, bots and headless browsers that load pages and, often, ads. Advertisers don't want to pay for impressions served to a server, and industry definitions treat known data-centre traffic as invalid. For a publisher, it's the easiest category to justify filtering.

VPN traffic

VPNs are more mixed. Plenty of real people use them for privacy, for work, or because their company network routes through one. Blocking every VPN visit would remove genuine readers.

But VPNs are also used to disguise bots and to fake a location, for example to make traffic look like it comes from a high-paying country. The sensible approach is to treat VPN as a risk signal, not a verdict: combine it with other checks and look at the pattern. A trickle of VPN visits behaving normally is fine. A wave of VPN visits from one provider, all landing on the same pages, isn't.

Proxy traffic

Proxies route traffic through another machine. The tricky kind is the residential proxy, which sends traffic through ordinary home connections so it looks like it comes from a real household. These are widely used to hide automated traffic, and they're harder to spot than data centres because the IP address itself looks normal.

Catching them usually needs more than an IP list: inconsistencies between the claimed location, the device and the browser, or behaviour that doesn't fit a human.

Why location faking matters

Ad prices vary a great deal by country. That gives fraudsters a reason to route traffic through proxies and VPN exits in high-value markets, so cheap visits look expensive. A useful check is to compare signals that should agree: the IP address location, the browser language, the device time zone and the time of day people visit. Real readers in a country mostly look consistent. Disguised traffic often doesn't: a visit that claims to be in one country while the device clock says another, at an hour when that country is asleep, at steady intervals. None of these proves anything alone, but together they tell a clear story.

What it looks like in practice

Across the sites Bidora manages, one 24-hour IVT Shield snapshot measured about 150,000 visits. Around 134,100 were estimated human, the overall invalid-traffic rate was 8.0%, and 11,600 ads were withheld. The interesting number was the outlier: one site that day was at 40.7% invalid traffic. Averages hide that kind of problem, which is why we look at each site on its own.

How to handle each type

SourceRiskSensible response
Known crawlersLow, but not ad-worthyLet them read; don't serve ads
Other data-centre visitsHighWithhold ads; block if volume is heavy
VPNMixedCombine with other signals before acting
Residential proxyHigh when detectedWithhold ads; investigate the source

The key idea is to withhold the ad rather than block the reader. That keeps your content open to everyone while protecting advertisers and your account. If you see a pattern you can't explain, our guide to signs of bot traffic walks through how to investigate.

This is the approach behind IVT Shield: every visit is classified as human, known crawler or suspicious, each verdict comes with a reason, and suspicious visits don't get ads.

Talk to us about joining Bidora MCM. We send Manage Account invitations, Bidora's share starts at 8%, and IVT Shield is included for every site we monetise. Start on the Get started page or email partnership@bidora.one.

Ready when you are

Tell us about your site. We'll take it from there.