Half of your digital advertisement budget may be buying ghosts
Wasted digital ad spend in India runs into tens of thousands of crores each year, sustained by bots that scroll, hover and 'watch' ads just long enough to get paid.
A large portion of advertisements on websites and apps—about 30 -40%—are never seen by human eyes, never processed by human minds, and never shown to real users at all.
According to mFilterIt, a digital trust and intelligence company that analysed 342 advertising campaigns in 2024, the industry still treats "viewability" as proof of success.
Under current standards, an ad counts as "viewable" if at least 50% of its pixels appear on screen for one second (two seconds for video). This threshold determines billing. Advertisers pay when this technical condition is met.
But viewability only measures whether an ad could have been seen—not whether anyone actually looked at it, processed it, or even existed on the other end.
These ads are served to bots. Displayed for milliseconds on screens no one is watching. Loaded beneath content while users scroll past. The ads technically "deliver." The platforms bill for them. But no human sees them.
The gap between visible and viewed
Consider this scenario: someone is watching a cricket match on their phone. An ad loads at the bottom of the screen and sits there for one second. It meets the viewability threshold. The advertiser gets billed. Then the viewer flips to WhatsApp.
No one saw the ad. The company still paid for it.
"You can see from mouse movement and behavior when someone flips away, but the system still counts it as a view," Amit Relan, CEO and co-founder of mFilterIt explained in an interview with YourStory.
This gap—between technical visibility and actual human attention—is where fraud operates. According to mFilterIt's analysis, between 30% and 45% of "valid" programmatic traffic fails what the industry calls "deep validation." These are impressions that exhibit behavior patterns inconsistent with actual humans browsing content.
Bots are now designed to pass the one-second test. They scroll. They hover over content. They watch videos for exactly two seconds, then move on. They mimic human boredom and distraction. They clear billing thresholds without being human.
“Earlier, the measurement metric was that viewable means visible, but a machine can also do this. A bot can also do it, so the question is How do you measure real human attention?” Relan said.
This also affects retargeting, which means showing advertisements to people who visited your site. It is efficient because brands are reaching ‘warm’ prospects who have already expressed interest.
However, if 30% of your ‘original’ site visitors were bots, then retargeting would simply mean a brand’s money is going towards showing advertisements again to these bots.
They are soaking up impressions.
While one impression may cost little. It quickly adds up. “The wastage is anywhere between 12 to 14% of digital marketing and in India alone, we are talking about ten to twelve thousand crores annually, and globally, it is almost a hundred billion dollars,” Relan said.
The app downloads that never happened
App install campaigns are worse. Companies pay Rs 150 to Rs 500 per install, depending on the category.
While things appear to be running smoothly on dashboards, retention data shows otherwise.
For example, if a company sees 10,000 installs a month, the retention rate might be as low as 40%, which means a majority of users never opened the app a second time.
Some of them were never users. They were emulators—software farms that generate fake installs at scale. Each one getss a synthetic identity: fake device ID, fake location, fake behaviour.
The install registers as real. The attribution platform credits the publisher. Companies pay for the “users” that were never there.
“Anyone can create synthetic identities and generate fake installs, and that entire money gets drained without real users,” Relan said.
This is not rare. According to mFilterIt's analysis, 43% of invalid traffic comes through affiliate networks, where install fraud is most concentrated.
Affiliate networks are platforms that connect merchants (brands or advertisers) with publishers (affiliates like bloggers, influencers, or website owners) to facilitate performance-based marketing.
Sites built to waste your money
Then there are made-for-advertising sites. These are not real publications. They exist only to sell ad space.
The content is garbage—usually AI-generated articles scraped from other sites, rewritten just enough to avoid copyright filters.
The layout is designed to load as many ads as possible, not to be readable. A visitor lands on the page, triggers ten ad calls, and leaves. Companies book for these ad slots but nobody reads these ads because no human visited the site.
“These are made-for-advertiser sites and AI slop sites that exist only to generate impressions and nothing else,” Relan said.
Some of these sites go further. They use auto-refresh scripts that reload the page every few seconds without the user doing anything, generating new impressions each time. Or they stack ads in invisible frames off-screen but report them as viewed.
Poor targeting
It is not just about bots; it is about bad placement. Algorithms often optimize for the cheapest views rather than the most relevant ones.
In a striking finding from mFilterIt’s analysis, 7% to 9% of YouTube impressions in analyzed campaigns ran on "Made-for-Kids" content. For a brand selling insurance, cars, or enterprise software, this is wasted on an audience that cannot convert, yet the system marks it as a successful delivery.
Companies set campaign frequency caps—showing an ad only a few times to a user—to prevent ad fatigue.
However platforms report average frequency. This means that while a company might have frequency cap set at five, one user might see an ad once while the other might see it 20 times. The average still remains five, rendering the cap meaningless.
“Wherever the money is, that’s where the anomalies will always exist, and that’s what needs to be measured and controlled,” Relan said.