In the days of print, advertising was simple. Sure, a magazine could inflate its readership on the margins, but brands could nonetheless be confident that their ads would be seen by actual human beings, as opposed to a swarm of malicious bots.

Those golden days are, of course, long over. Today’s marketers, plying their wares online, are operating in an environment rife with ad fraud, on a sometimes staggering scale: in recent years, over a fifth of all digital ad spend was attributed to fraud, amounting to $84 billion.

It’s true that that number would likely be significantly higher without aggressive ad-marketplace reform efforts over the last decade. But it is not enough to simply say “could be worse” and leave it at that. Not when marketers are losing millions, and especially not when they don’t have to be.

Despite reforms, fraud remains rampant

Anti-fraud initiatives like those undertaken by the IAB Tech Lab have chipped away at this problem without coming close to solving it. We can see this quite clearly through the example of connected TV (CTV), perhaps the fastest-growing ad sector in recent years, and also the sector most plagued by fraud.

In 2020, for instance, the Icebucket scam impersonated millions of “viewers” for high-CPM video ads, in what some have suggested might be the biggest ad fraud scheme in history. At its peak, the scam accounted for nearly a third of total volume on programmatic platforms for CTV, a feat made possible through the counterfeiting of over 300 fake publishers.

Similar scams have proliferated in the years since, from the ParrotTerra scam in 2021 (in which advertisers came close to losing $50 million before detection) and SneakyTerra. These of course, come on the heels of a number of well-publicized scams in the world of conventional programmatic advertising, most prominently the 3ve Botnet, which cost businesses $29 million in ad spend pre-detection.

In addition to affecting a business’s bottom line, large-scale deceptions like these affect the security of customers. And AI is only turbocharging the problem: see, for instance, the AI-powered ad-click malware attack identified by researchers this past January. Here, malware would launch a hidden browser in the background of a user’s phone, load ads invisibly, and then simulate scrolling and clicking. While (to our knowledge) this only led to lost ad spend, this same technique could easily be used for more malicious payloads.

Meanwhile, the advent of agentic AI makes all of this infinitely more complex. Previously, marketers could identify ad fraud by tracking the behavior of bots, which tended to be much more rigid and mechanical than that of humans. But the behavior of agents is much more human-like: unlike bots, they might pause while scrolling or even abandon active carts. Consequently, pinpointing bot activity has never been more challenging.

Marketing Technology News: MarTech Interview with Theresa Pham, Head of Product @ Wayvia

How to fight ad fraud

Clearly, large-scale, collaborative efforts are a prerequisite to meaningful change on this front. But there are absolutely precautions that individual companies can take to prevent wasted ad spend in the here and now.

If advertisers consistently monitor campaign performance, fraudulent activity becomes much easier to identify and eliminate. Click-based ad fraud that generates engagement without real conversions can be quickly flagged and removed, especially when digital ad platforms automate this process.

Even more sophisticated fraud schemes that simulate conversions can be uncovered when advertisers connect their ad campaigns to CRM systems, order data, and actual purchase or subscription metrics, allowing advertisers to distinguish genuine customer activity from fabricated results.

Most significantly, these include:

1) Click origins:

If 30% of your clicks are suddenly coming from a country that doesn’t match your customer base, that’s a fairly reliable indication that your traffic is fraudulent.

2) Speed of access:

If you notice the same device signature accessing two or more campaigns across different platforms within seconds, you can be fairly confident you’re dealing with automation. This connection can’t really be made on the platform end, because ad platforms only see their own traffic: it’s up to the marketer to suss out the potential fraud.

3) Be mindful of return on ad spend:

One thing a bot will almost certainly never do is actually buy one of your products. In other words, while clicks are eminently fakeable, actual sales are much less so. It’s not uncommon for some marketers to get lost in the data and lose track of the bottom line. This is a tendency that is very much worth resisting.

Some degree of fraud in the online ad space is probably inevitable. But the degree of fraud we’re seeing now is not. As exciting as the advent of AI has been for marketers, it has unleashed perils whose scope we are only now beginning to understand in full. They are sweeping, complex, and ignored only at marketers’ peril.

About Shirofune

Shirofune, an AI-driven digital advertising automation platform used by major global brands and agencies (such as Dentsu) to optimize campaigns across platforms like Google, Meta, and TikTok