The Invisible Media Problem

The Invisible Media Problem

Digital advertising has perfected the science of buying impressions. The trouble is that impressions are not the same thing as human attention.

Back when the media conversation was more simple, we optimized for reach and frequency, negotiated the cost of a page in The Wall Street Journal, debated whether the evening news was worth the CPM. Media buying felt like placing chips on a table and betting on the audience seated directly across from you.

Throughout my career sitting in agency and CMO chairs, I have overseen hundreds of millions of dollars of media spending. The fundamental question was always the same: how do we persuade people?

Television delivered mass reach. Print delivered credibility. Outdoor delivered frequency. None of it was perfect but at least you knew the audience was there. You might not have known exactly which living room the message landed in, but you knew it had landed in an actual living room, not in a server farm in Latvia talking to a bot pretending to be a man comparing trainers.

Then programmatic arrived.

Suddenly media buying was no longer about audiences but about inventory. Dashboards replaced judgment. Algorithms replaced media planners. The language shifted from persuasion to optimization. Somewhere along the way we stopped asking a much more basic question: is anyone actually seeing this?

Today digital media has reached a strange place. We have built the most sophisticated advertising infrastructure in history, but increasingly it appears to be optimized for impressions that humans barely notice.

Which brings us to what I call the invisible media problem.

Half the Internet Isn’t a Person

One of the realities of modern marketing is that a massive share of the internet is no longer human.

According to Imperva’s 2024 Bad Bot Report, 49.6% of all global internet traffic is automated and 32% is classified as malicious bots. That means roughly half of what moves across the web is not a person.

This matters because automated traffic does not just scrape websites, test passwords or hoover up content for some AI model with delusions of grandeur. It also touches the advertising ecosystem.

The Association of National Advertisers estimated that ad fraud cost marketers $22 billion globally in 2023.

Even in well-managed campaigns, fraud detection firms like DoubleVerify and Integral Ad Science regularly report 10–15% of impressions failing quality checks such as invalid traffic, fraud or unsafe environments.

So before we get to creativity, persuasion or brand storytelling, we have a more awkward issue to address: a measurable portion of the world’s advertising budget is effectively negotiating with software.

The Ad-Farm Internet

But bots are only part of the problem. A much larger issue is the explosion of websites created primarily to harvest advertising.

In 2023 the Association of National Advertisers conducted a major study into Made-For-Advertising sites. These are web properties designed almost entirely around ad placement rather than editorial value. The study found that 21% of programmatic advertising spend flowed to these sites.

They are easy to recognize once you know what to look for: endless slide shows, recycled articles, celebrity gossip scraped from other publications, AI-generated filler and a constellation of banner ads orbiting the page like satellites. They are less media brands than industrial sheds for storing ad units.

Brands such as Procter & Gamble and Unilever have both publicly pushed agencies and media partners to reduce spending on these environments after discovering how much budget was quietly drifting into them. Unilever’s former CMO Keith Weed warned years ago that the industry risked funding what he called a swamp of content.

The swamp has only grown. Now it has algorithmic drainage, synthetic articles and better monetization.

 Procter & Gamble cut over $200 million from digital advertising after finding waste and questionable placements, with little impact on sales.

Seen but Not Really Seen

Even when ads reach legitimate websites and real people, another problem emerges: attention.

Research from Lumen Research, which tracks eye movement and attention in advertising environments, shows that many digital display ads receive less than one second of human attention. The advertising measurement firm Adelaide has found that a majority of programmatic impressions generate minimal measurable attention time, often fractions of a second. Meanwhile the Media Rating Council’s viewability standard allows an ad to count as viewable if 50% of its pixels are on screen for just one second.

One second.

An ad can technically be counted as seen even if a human barely registers it. That may satisfy a dashboard, but it would not satisfy any sane client if you described it honestly. “Good news, your ad was visible for roughly the duration of a hiccup.”

This is the digital equivalent of paying for a billboard that drivers glimpse only while sneezing. 

The industry has become dangerously comfortable confusing the opportunity to be seen with actual seeing.

The Smart Brands Are Already Cleaning House

Some of the largest marketers in the world have begun questioning the system and, more importantly, acting on it.

Procter & Gamble, one of the world’s biggest advertisers, cut more than $200 million from digital advertising budgets after discovering waste and questionable placements. According to the company, the reduction had little negative impact on sales.

That should have set off more panic than it did. When a giant removes that much spend and the sales line barely flinches, it is not a trimming exercise. It is an indictment.

This is the digital equivalent of paying for a billboard that drivers glimpse only while sneezing. 

Chase Bank conducted its own brand safety audit and discovered its ads were appearing on 400,000 websites through programmatic buying. After tightening controls the bank reduced placements to around 5,000 sites with no measurable loss in performance.

Think about that for a moment. Four hundred thousand sites to five thousand. Same ballpark of results. That’s like discovering you have been buying the digital equivalent of air.

Meanwhile Mars, the global food company behind brands like M&M’s and Snickers, has been outspoken about pushing the industry toward attention-based media measurement, arguing that impressions alone are increasingly meaningless.

When companies spending billions begin questioning the system, it usually means something deeper is happening.

Chase reduced programmatic placements from 400,000 websites to about 5,000 to realize there was no measurable loss in performance.

AI Is About to Flood the Market With More Junk

If the invisible media problem feels large today, artificial intelligence may soon multiply it. Generative AI tools have dramatically reduced the cost of producing content at scale. Entire websites can now be created automatically, filled with machine-generated articles designed primarily to capture search traffic and advertising revenue.

A 2024 analysis by NewsGuard identified hundreds of AI-generated news sites appearing across multiple countries, many operating with minimal editorial oversight but fully monetized with advertising.

Meanwhile the Reuters Institute Digital News Report found that around 40% of global consumers say they struggle to distinguish AI-generated content from human journalism.

If consumers can’t reliably distinguish real from synthetic and marketers can’t reliably distinguish quality environments from monetized rubbish, the system becomes more efficient at distribution and less trustworthy at every level.

Cheap content plus automated advertising infrastructure creates the perfect conditions for a new generation of invisible media environments.

In other words, the supply of places to put ads is now growing faster than the supply of real human attention. Marvelous for the ad-tech ecosystem. Less than marvelous for anyone hoping to persuade an actual buyer.

Closed ecosystems built on real user identities give platforms like Google, Meta and Amazon a certainty the open web cannot match.

The Ad-Tech Tollbooth Problem

Another twist in the story is that brands increasingly rely on complex ad-tech supply chains to reach audiences.

The World Federation of Advertisers has warned that the programmatic ecosystem can involve dozens of intermediaries, each taking a small share of media budgets while adding opacity to where ads actually appear.

A landmark ANA programmatic transparency study found that advertisers often see only about 60 cents of every programmatic dollar reach working media, with the rest absorbed by fees across the ad-tech chain.

No marketer would tolerate losing forty cents on the dollar in any other part of the business. Imagine a CFO being told, “Excellent news, we have located a system where nearly half your money disappears before the work begins.” In most departments that would end in handcuffs or at least a tense board meeting.

Yet in digital media it has quietly become normal, partly because the machinery is complex and partly because complexity has an unfortunate habit of dressing up waste as sophistication.

Why the Platforms Keep Winning

Ironically the invisible media problem has strengthened the largest platforms.

Because they operate closed ecosystems with real user identities, companies like Google, Meta and Amazon offer advertisers a degree of certainty that much of the open web cannot match.

According to GroupM’s global advertising forecasts the largest digital platforms now capture well over half of global digital advertising spending.

Retail media networks have grown particularly fast. Amazon’s advertising business surpassed $46 billion in revenue largely because brands know their ads appear directly next to products in environments where shoppers are actively making decisions.

That is the crucial distinction. Amazon may charge handsomely, Meta may never stop inventing new acronyms and Google may continue its charming habit of changing the furniture every six minutes, but at least marketers can see the commercial logic. The ad is near intent. The buyer is plausibly real. That’s a whole lot better than an AI-generated article called “Seven Surprising Facts About Dishwashers” floating on a site nobody has ever visited on purpose.

When advertisers doubt the open web, they migrate toward environments where attention is easier to verify.

What Marketers Need to Do Now

First, shift the metric from impressions to attention. If an ad is technically served but barely noticed, that’s not media value. 

Second, audit where your ads actually appear. Not in theory. Not by category. Actually appear. Many brands discover astonishing things when they do and few of those discoveries are delightful.

Third, limit exposure to Made-For-Advertising (MFA) sites. A surprisingly large share of programmatic spend flows to these properties by default because the system is built to find cheap inventory before it finds meaningful context.

Fourth, prioritize context over scale. Advertising next to credible journalism, respected creators and trusted platforms often produces far stronger results than chasing cheap inventory across the open web. A smaller audience paying real attention beats a vast digital wasteland every day of the week.

Fifth, reduce supply-path complexity. If too many intermediaries are touching your money before your message reaches a consumer, your budget is not working hard enough.

Sixth, demand human outcomes from machine systems. Programmatic is a tool, not a theology. Optimization is useful, but when every decision is driven by algorithms chasing the lowest CPM the system inevitably finds the cheapest attention available. And cheap attention is often invisible.

Seventh, rebuild media judgment as a strategic skill. The industry became so enchanted by automation that it started treating discernment as old-fashioned. It is not old-fashioned. It is what prevents nonsense from being purchased at scale.

The Question That Ruins the Presentation

After decades in marketing I have learned one small but useful trick. Whenever someone presents a digital media report full of impressive numbers, I ask a very old-fashioned question: Where did the ads actually appear?

That question has an almost magical ability to suck the oxygen out of a room. Suddenly the conversation moves from abstractions to accountability. Because once that question is asked the invisible media problem stops being theoretical.

It becomes a judgment problem, a procurement problem, a governance problem and, for any marketer with responsibility for real money, a professional pride problem.

Media was supposed to help persuade people. Somewhere along the way too much of digital media became a machine for serving ads into murky environments, reporting back numbers and calling the whole thing progress.

It’s not progress if nobody sees it. It’s not efficiency if nobody notices it.
And it’s certainly not brilliant targeting if your so-called best-performing audience turns out not to be a middle class family from East London but a robot family from a server farm East of Bucharest.

The invisible media problem is not that digital advertising no longer works. It’s that too much of the system has become indifferent to whether it works for the reason we pretend it does. That is the plot it has lost.

Sources: Association of National Advertisers Programmatic Media Supply Chain Transparency Study; Association of National Advertisers Made-For-Advertising Website Study (2023); Imperva Bad Bot Report (2024); Lumen Research Attention Measurement Studies; Adelaide Attention Metrics Research; Media Rating Council Digital Viewability Standards; NewsGuard AI Generated News Site Investigation (2024) ; Reuters Institute Digital News Report (2024); World Federation of Advertisers Programmatic Supply Chain Analysis; GroupM Global Advertising Forecast.

John Rose

Creative director, author and Rose founder, John Rose writes about creativity, marketing, business, food, vodka and whatever else pops into his head. He wears many hats.