ILLUSTRATION: Oliver Hazelwood/Shutterstock
Disclosure: Longtime journalist Steven Brill is the founder or cofounder of a number of publications and companies, including NewsGuard, where he is the co-CEO and coeditor in chief. Among other services, NewsGuard offers advertisers brand-safety services aimed at countering the pitfalls of unvetted programmatic advertising. This story is excerpted from his new book, The Death of Truth.
In 2019, other than the government of Vladimir Putin, Warren Buffett was the biggest funder of Sputnik News, the Russian disinformation website controlled by the Kremlin. It wasn’t that the legendary champion of American capitalism had an alter ego who woke up every morning wondering how he could help finance Vladimir Putin’s global propaganda network. It was because Geico, the giant American insurance company and subsidiary of Buffett’s Berkshire Hathaway, was the leading advertiser on the American version of Sputnik News’ global website network.
Nor was it because a marketing executive at Geico had decided that advertising on the Russian disinformation outlet was a good idea. That would have been especially unlikely, not only because of the Buffett connection, but also because Geico stands for Government Employees Insurance Company and has its roots dating to the 1930s, providing insurance to civilians and members of the military who worked for the American government, not its Russian adversary.
In fact, no one at Geico or its advertising agency had any idea its ads would appear on Sputnik, let alone what anti-American content would be displayed alongside the ads. How could they? Which person or army of people at Geico or its agency could have read 44,000 websites?
Geico’s ads had been placed through a programmatic advertising system that was invented in the late 1990s as the internet developed. It exploded beginning in the mid 2000s and is now the overwhelmingly dominant advertising medium. Programmatic algorithms, not people, decide where to place most of the ads we now see on websites, social media platforms, mobile devices, streaming television, and increasingly hear on podcasts. The numbers involved are mind-boggling. If Geico’s advertising campaign were typical of programmatic campaigns for broad-based consumer products and services, each of its ads would have been placed on an average of 44,000 websites, according to a study done for the leading trade association of big-brand advertisers.
Geico is hardly the only rock-solid American brand to be funding the Russians. During the same period that the insurance company’s ads appeared on Sputnik News, 196 other programmatic advertisers bought ads on the website, including Best Buy, E-Trade, and Progressive insurance. Sputnik News’ sister propaganda outlet, RT.com (it was once called Russia Today until someone in Moscow decided to camouflage its parentage), raked in ad revenue from Walmart, Amazon, PayPal, and Kroger, among others.
Every workday, approximately 2,500 people sit at desktops or laptops using these programmatic advertising algorithms to spend tens of millions of dollars an hour. They work at advertising agencies scattered around the world, or, in the case of some major companies, at their in-house advertising shops. Their titles might be “programmatic specialist,” “programmatic associate,” or “campaign manager.” What they have in common is that they are usually in their first jobs out of college. Although many work from home post-Covid, if they are in the office, they sit at carrels in large open spaces that resemble the trading floor of a stock brokerage.
A Keyboard Replaced Mad Men
Let’s call our archetype specialist Trevor, and assume that he works in the programmatic advertising unit of one of the five major global advertising agency holding companies. He probably has a salary of $60,000 to $80,000 a year. Trevor will be logged in to what is known as a demand-side platform. Think of it as a kind of stock exchange for buying advertising instead of shares of a company. The demand-side platform is where all of the available advertising space on every page of every website in the world that the platform has assembled as its inventory is made available to a buyer like Trevor.
In proximity, or in close touch if working remotely, will be another junior staffer with a title of “media buyer,” “planner,” or “campaign manager,” whose job is to make sure that the advertising effort, or “campaign,” that has been planned by higher-ups on the creative and planning teams is communicated to Trevor. This includes loading the actual ad for the product onto the demand-side platform for deployment, and also giving Trevor, sitting in front of the demand-side platform’s dashboard, the all-important targeting decisions that the planners have made: Who should be reached with what message? Yes, humans are still involved in picking the sales strategy and creating the message (although generative AI may change that, too). However, humans do not decide which publisher—the local newspaper website, or a website posing as a local news site but publishing Russian propaganda—gets the ad.
Trevor will then go through a series of screens offering an array of choices for where, how, and when the ad will appear. The most important and complicated set of choices has to do with reaching the ad’s target audience at the best price.
The process is like a stock exchange, except that the buyer doesn’t know what stock he is buying, meaning that the advertiser doesn’t know whose advertising inventory he is buying.
What Trevor is not offered is a choice of specific places to advertise. There are planners in a smaller, different department at his agency who still do what is called direct buys—choosing specific newspapers, magazines, or websites where they want an ad to appear. But they are a dwindling minority, because direct buys are fast becoming a relic of a different time. Trevor’s keyboard has replaced the marketers of the 1960s Mad Men era, who decided over lunch and cocktails whether to advertise in Time instead of Newsweek, in Businessweek instead of Fortune, or on NBC instead of ABC. Almost all advertising online—and even much of it on television (through streaming TV), or on podcasts, radio, mobile devices, and electronic billboards—is now done programmatically, which means the machine, not a planner, makes those placement decisions. Unless the advertiser uses special tools, such as what are called exclusion or inclusion lists, the publishers and content around which the ad appears, and which the ad is financing, are no longer part of the decision.
Trevor’s targeting choices start with obvious variables and then can become almost infinitely granular, offering a stunning display of the depth of data that has been collected about all of us:
Click for male or female.
Click for age-group.
Click for where the target lives, down to zip code or classification of zip code (urban, suburban, wealthy, middle class).
Click for income level.
Click for education level.
Click for affiliation with a religion.
Click to limit the number of times the target will see the same ad.
Click for device used. (Is the person going to see the ad on a mobile device, on a laptop, on a streaming video service?)
Click for time of day you want the ad to run. (The programmatic specialist wouldn’t want a McDonald’s breakfast ad to run in the evening.)
Click for an “intent signal.”
That means some recent behavior available in the data (such as someone having visited a type of website) that indicates a propensity to:
Buy a pickup truck.
Join a gym.
Take a vacation.
Buy broadband service or an alarm system for a new home. (This might come from data available from change-of-address forms filed with a utility.)
Rent a car.
Buy a baby crib.
Order dinner to be delivered.
Be looking for a new job. (Has the person visited a job recruitment website recently?)
Buy a car.
Have breakfast at a fast-food restaurant.
Shop for diabetes medication.
Buy pet food.
Vote for a particular political candidate.
There are hundreds of such categories of intent signals. And there are “next” boxes that Trevor can click on to go still more granular, such as picking the brand of car that the person has shown an interest in buying. Or he might click only a few or none of them, and move on to the next set of variables if, for example, the brand is a widely used consumer product, such as Coca-Cola, for which some of this granular targeting may not be relevant.
Having checked all the right boxes and clicking “next” to go through all the pages with all the choices that move across his screen, Trevor has picked his target audience. Perhaps it’s a high-income, college-educated male aged 35 to 65 who lives in the US suburbs, has shown a propensity to rent cars, travels regularly by air on business, stays at high-end hotels, and has shown interest in electric cars. Now Trevor has to bid on what the client will pay to reach these targets.
His client—let’s say it’s Hertz—has allocated $100,000 for a campaign to boost its US business by showcasing the premium e-cars it has available. Trevor decides to bid what is called a 20-cent CPM—20 cents for each view of the ad among 1,000 people in the target audience reached over the next six months. The demand-side platform tells him there are 1 million such “targets” currently available for bidding. At 20 cents per thousand, it will cost $200 to reach all of the targets once, and it will cost $100,000 (his budget) to reach the targets 500 times over the six months. In the advertising business this is known as 500 impressions. So, Trevor is buying 500 million impressions (500 × 1 million targeted people). That’s the plan. Another part of the plan is to space the advertising out evenly over 180 days. He has not expressed a preference for what time of day the ad will be viewed.
However, Trevor may end up saving money. He has “bid” 20 cents per thousand. That is the maximum he will pay. But when the demand-side platform canvasses its inventory of ad spaces available on tens of thousands of websites, it may find that some websites (or networks of websites, called supply-side platforms) will demand only 5 cents per thousand. The price may be depressed because, at the moment Trevor’s bid goes out, the suppliers have lots of unsold inventory, or advertising space, including inventory about to be viewed by some of Trevor’s targets. So, the algorithm allows the suppliers to offer a bargain instantaneously so that they can unload what will otherwise be unsold ad space on their websites. Trevor can then decide to take the savings back to his client or increase the number of impressions he buys.
Amazingly, the algorithms created and put in place over the past decade are such that this pricing of the ad—website by website, target viewer by target viewer—happens in a fraction of a second. Each of Trevor’s 500 million impressions, or ads, will be bought separately, one by one, through this near-instantaneous auction system.
Trevor and his planning partner are likely to have a second screen where they can monitor the data that the demand-side platform will send them minute by minute over the next 180 days, demonstrating how the campaign is doing: the price Trevor is actually paying, the campaign’s progress in reaching the targets, and the results, including the percentage of targets who are clicking through to get more information from Hertz. This allows them to adjust the plan—reduce or add to the budget, tinker with the message, or change some of the target choices.
You Are the Product Being Sold
Spending on programmatic advertising globally is estimated to reach more than $300 billion in 2023. It involves a chain of commerce, starting with the advertiser and ending with the publisher of the ad, that is far more complex than the oversimplified version I have just tried to walk you through. In fact, it is so complex and so opaque that when pressed, most if not all of those who have been immersed in the industry for more than a decade do not understand all of the jargon or every aspect of how it works, or even how effectively it works. Yet what I kept hearing as the professionals explained it to me was that the process is like a stock exchange, except that the buyer doesn’t know what stock he is buying, meaning that the advertiser doesn’t know whose advertising inventory he is buying. That’s right: The advertiser and its ad agency have no idea where among thousands of websites its ad will appear.
That may be true, but it misses the most important point, which is that the publisher and the publisher’s content are not the product. The product is you—the person whose data has been harvested so exquisitely that you are the advertiser’s target. You are the “stock” that the advertiser is buying. The core idea of programmatic advertising is that where you are seeing the ad doesn’t matter.
This marks a sea change in the advertising industry. Until the mid-2000s, publishers employed legions of salespeople to convince those with products or causes to sell that their pages or TV or radio shows were the right environment for their ads. The content counted. With the coming of programmatic advertising, the content of the tabloid National Enquirer compared with that of the reputable Philadelphia Inquirer didn’t matter. What mattered was where you were likely to get your target to look at your ad for the lowest price.
There are multiple arguments contradicting the assumption that where ads run makes no difference, including studies showing that people respond more positively to advertising that appears on websites and in other media that they take seriously. Yet programmatic advertising has thrived based on the central belief that all impressions aimed at the right target are equally valuable. So, if Sputnik News is selling an impression for less than a legitimate local newspaper is asking for it, Sputnik will win the auction. It’s a perpetual, instantaneous race to the bottom. If the bid for an impression on the Santa Monica Observer—a hoax website that ran a phony story about Nancy Pelosi’s husband, Paul, being with a male prostitute when he was brutally attacked last year—is lower than the bid offered for an ad on an article that tells the real story of what happened to Pelosi published by the San Francisco Chronicle, which pays real reporters to write real stories, then Hertz’s ad will be on the Observer story. As, indeed, it was.
The (Programming) Bonanza
That same 2023 study by the advertiser trade association that estimated the number of websites involved in a typical big-brand campaign also focused on the quality of those websites. It dubbed websites with clickbait headlines and stories like the one about Paul Pelosi as MFAs, or made-for-advertising sites, meaning that their only purpose is to get on the programmatic advertising gravy train using whatever headlines, articles, and images work best to attract the social media likes and retweets that will lure readers to the site so that they see a programmatic ad. The study found that the Trevors of the world end up spending 14 percent of their ad dollars on MFAs, which would be $42 billion.
The two biggest demand-side platforms responsible for making this multibillion-dollar, multi-buyer, multi-seller auction happen are Google and the Trade Desk, a company based in Ventura, California. Google enjoys the dominant share of the demand-side market. At number two, the Trade Desk does about half the business Google does, if that.
Google does not disclose much about its volume and profit, and because its demand-side platform, while huge, is only a part of Google’s overall business, the company does not have to break out the details in the reports that publicly traded companies are required to file. More can be known about the Trade Desk and the economics that make programmatic advertising so compelling. Like Google, it is a publicly traded company, but its demand-side platform is its only business, which makes its filings a window on at least some of the details of the programmatic advertising business model.
The Trade Desk was founded in 2009 by two recently departed Microsoft employees. Its cofounder and CEO, Jeff Green, had a net worth of $4.2 billion from his stock in 2023, according to Forbes. In 2022, the company reported revenue of $1.578 billion, an increase of 32 percent over the prior year. Because it is so AI- and data-centric, in 2023 it needed only 2,800 employees. WPP, the largest ad agency holding company, has more than 100,000 employees. Less than 3 percent of that is all it takes for the Trade Desk to create the algorithms that manage and massage data from its own data bank of online sources—as well as from dozens of outside data brokers that mine credit reports, license applications, and census counts—and to guide clients on how to use it all. As a result, the Trade Desk’s 2022 cash flow (profit not counting accounting adjustments) was a jaw-dropping 42 percent of revenue, or $668 million. Its market capitalization (the value of all of its stock) was about $38 billion as of September 2023. WPP’s market value was about a fourth of that: $10 billion.
In a pending antitrust complaint filed in 2021 against Google’s programmatic advertising arm (that remained pending as of the end of 2023), lawyers for the Texas attorney general described the daunting numbers involved in the demand-side platforms’ handling of all that data:
One might think that a website with three pages and three different ad slots per page would have a total of nine unique ad units to sell. But because online ads are targeted at individual users, the same site with 1,000,000 readers actually has 9,000,000 different units to sell: each of the website’s impressions targeted to each unique reader. Consequently, an online publisher’s inventory is akin to the inventory of seats at a baseball stadium. No two seats are exactly the same.
This analogy of no two seats in the stadium being exactly the same is why, if you and I go to the same website at the same time, we will see two different ads depending on what demographic data, location data, intent signals, and other indicators are linked to each of our devices.
The Trade Desk’s 2019 annual report explained what all that technology is doing behind the scenes as Trevor pushes the buttons on his console:
“On average, our real-time bidding technology evaluates more than 790 billion ad opportunities per day, reaching over 819 million devices per day on a global basis … We use the massive data captured by our platform to build predictive models around user characteristics, such as demographic, purchase intent or interest data. Data from our platform is continually fed back into these models, which enables them to improve over time as the use of our platform increases … Our bidding engine then shifts bids and budgets in real-time to deliver optimal performance.”
Evaluating “more than 790 billion ad opportunities per day” means that the Trade Desk machine is assessing more than 5.4 million ad impressions per second. If Trevor is using the more dominant Google instead of the Trade Desk, those numbers are likely at least to be double.
What the Google or Trade Desk machine is not doing is telling Trevor where those impressions are—which 1,000 or 10,000 or 50,000 (again, the average is 44,000) websites his ad is going to.
The ‘Blocking Words’ Sledgehammer Solution
As programmatic advertising picked up steam, the industry realized that being totally content agnostic about where ads appeared was a problem. Ads were appearing on pornography sites, on sites promoting racism or antisemitism, and on sites that promoted terrorism or recruited terrorists. Other sites used bots to create fake viewership; the bot would make it seem as if a real person were viewing the site, thereby boosting the number of ad impressions it could sell.
As a result, between 2008 and 2010, three companies were launched to capture what became a new “brand safety” market—the business of keeping products from companies like Coca-Cola and Procter & Gamble brand-safe by keeping their ads off these brand-unsafe sites and rooting out fraud. Their technology was highly effective at spotting and blocking ads from ending up on porn sites and even some websites featuring particularly virulent hate speech or using common terrorism rhetoric; their algorithms, keyword search software, and other artificial intelligence tools could detect the images or words used on such sites. The software coders also developed ways to distinguish some bots from real people. Within 10 years, the three companies had become enormously valuable. Two went public at valuations exceeding $1 billion. The third was acquired by Oracle, the giant tech company, for $800 million.
But what artificial intelligence could not do was spot most forms of disinformation and misinformation, especially if the offending websites didn’t use obviously outlandish headlines or telltale provocative words or images. Soon, the three brand-safety companies developed a radically unnuanced, high-tech way to deal with the misinformation and disinformation problem: They offered advertisers the option to deploy “blocking words.” If any of these words appeared on any page of a website, then no ad from an advertiser or its agency paying the brand-safety company would appear on that page. The goal was to shield advertisers from content that was potentially brand-unsafe by deploying the companies’ software to avoid even the possibility of problem content.
Before programmatic advertising, brand safety had a far simpler and narrower meaning. If there was a plane crash, it was not considered brand-safe for an airline ad to appear next to a story about the crash. The publisher would take steps to make sure that did not happen or offer a refund if it did.
Now, as blocking words began to proliferate, the lists got longer and longer. They included any word—hundreds, even thousands of words, depending on the advertiser, ad agency, or brand-safety company deploying them—that those maintaining the lists thought might be a tip-off to an article that someone, somewhere would find controversial or unpleasant. “Shot” might indicate a story about a murder, which perhaps an advertiser might not want to appear next to, though no research has ever demonstrated that this would undermine an ad campaign. It might also be a story about a basketball game—as in, “He took a shot at the buzzer.” Either way, the ad was blocked from that page.
When Russia invaded Ukraine in 2022, the brand-safety companies failed to keep 79 brands owned by companies in Western democracies off 88 different Russian propaganda sites, a particularly ironic misdirection of advertising funds at the time.
When the Covid pandemic swept across the globe, these brand-safety companies added words associated with the pandemic to their blocking words lists, thereby offering a solution that was doubly toxic. Their artificial intelligence solutions could not keep ads off the disinformation and misinformation spreading about Covid if the sites spreading it were smart enough to avoid the blocking words. At the same time, these words effectively blocked advertising that would support reliable information in articles about Covid.
Yet to a brand’s marketing executives and its ad agency, it was a good way to keep from being embarrassed—or fired. Better safe than sorry.
A caveat: A major activity of NewsGuard has to do with selling itself as an alternative to blocking words and artificial intelligence when it comes to helping advertisers avoid having their programmatic ads run on egregious disinformation and misinformation websites, streaming television channels, or podcasts. Instead they can license our data, which identifies those meeting our criteria for adhering to the basic standards of journalistic practice, and then make informed decisions about how to use the data. Accordingly, I have a self-interest in persuading readers that NewsGuard offers a better “brand safety” alternative: human intelligence—actually reading and assessing news and information providers—rather than artificial intelligence. How the traditional brand-safety technology solutions performed during recent crises makes a good case for this alternative.
From when the Covid pandemic became a global headline in February 2020 through July 2021, 4,315 brands representing every kind of product bought more than 42,000 unique ads on websites flagged by NewsGuard for publishing Covid falsehoods.
The problem was global. In 2021, a French television documentary named multiple big brands—including the post office, the leading telecommunications company Orange, the government’s internal revenue service, and the retail giant Carrefour—that were helping to finance misinformation with their advertising.
At the same time, as the Covid crisis peaked, advertisers, their agencies, and the brand-safety companies that used blocking words took the practice to a new extreme. Because “Covid” itself became a blocking word, any article with the word could be blocked from having any ads. That, of course, meant that because Covid dominated the news and was featured or at least mentioned in multiple stories a day, much of the advertising inventory for the most reliable news sites was eliminated, and the most reliable reporting, including reporting that investigated and debunked misinformation about the virus, lost most if not all of the financial support that advertising provides. It got to the point where, on some days, nearly half the pages on the New York Times or the Wall Street Journal websites were filled with messages from low-rent advertisers paying pennies for ads because no big-name brand was buying them. Or the ad spaces on these online news pages were replaced with generic messages from the brand-safety companies, placed there to block the high-revenue ads that would otherwise have appeared.
The same meltdown happened in the period between the lead-up to the US 2020 presidential election and the January 6 riot at the Capitol. From October 1, 2020, through January 12, 2021, 1,668 brands ran 8,776 unique ads on the 160 sites flagged in NewsGuard’s Election Misinformation Tracking Center for publishing falsehoods and conspiracy theories about the election. At the same time, “election” and “Trump” were frequently used as blocking words, meaning that these sites either were getting ads from brands that did not deploy blocking words or had figured out how not to use the blocking words. When Russia invaded Ukraine in 2022, the brand-safety companies failed to keep 79 brands owned by companies in Western democracies off 88 different Russian propaganda sites, a particularly ironic misdirection of advertising funds at a time when so many companies in the West were otherwise publicizing their determination to stop doing business in or with Russia. And, once again, the list of words to be blocked was updated so that articles reliably reporting the truth about the invasion were stripped of financial support if they had a word like “invasion,” “Russia,” or “Ukraine.” Any of those words appearing on even the most trustworthy news site would mean that no ads from an advertiser using these blocking words would appear.
More recently, with the outbreak of the Israel-Hamas war, blocking words again proved to be an ineffective solution, either because many advertisers did not use them or because so many of the worst websites figured out how to evade them by not using the specific blocked words. Within two months after the war started, 349 top brands—including Macy’s, Zoom, Hulu, and the AARP—had been found advertising on websites promoting Hamas propaganda, including stories reporting that the Hamas October 7 attack was actually an Israeli “false flag” operation.
Among other keywords often blocked are “Black” and “gay.” This means that the many news publishers around the United States directed at the Black or gay communities have much of their news stories online deemed brand-unsafe, and thus lose the opportunity to generate much-needed ad revenue. This is happening at a time when many brands and ad agencies claim to want to increase their advertising to these communities. The publisher of PinkNews, a large gay-run news operation based in the UK, told me that there are days when most of his articles are deemed brand-unsafe, even though his readers trust the site to cover news topics of special interest to them.
The blocking words situation got so bad at one major global news publisher that executives assembled ad sales data into a chart that arranged revenue received by types of stories presented. Were they lighter features or hard news? Were they resource-heavy to report or resource-light? In the upper-right-hand quadrant were the revenue numbers for the news staff’s toughest-to-report stories on the most serious subjects. They were drawing only minimal revenue, even though their ads might have enticed readers at exactly the moment they were most engaged.
Still in Love With the Magic
Some veterans of the programmatic advertising boom have become cynical about the Frankenstein’s monster they have created.
Although no one knows for sure, the best estimates are that before the publisher gets paid for its ad, roughly half of every programmatic ad dollar goes to middlemen—the agency, the demand-side platforms, the supply-side platforms that assemble the publishers into networks to display for sale on the demand-side platforms, the high-tech brand-safety companies, companies like NewsGuard, and others with some piece of the action. Combine that skim off the top with ad fraud from bots and other schemes to defraud the machine, and the magic of the instantaneous auction to produce the lowest cost per thousand becomes less magical. Some advertising professionals told me that if the auction is based on the open internet rather than carefully curated networks of websites, it is less cost-effective than the old direct sales days depicted in Mad Men.
Nonetheless, the magic of programmatic advertising seems here to stay. In the United States, spending on ads delivered programmatically more than doubled from 2019 through 2022, to nearly $130 billion, which is about double the amount spent on all television advertising, national and local. As one senior executive at a major ad agency holding company explained, “We’ve created this giant multibillion-dollar machine. It produces higher margins for us than anything we could ever do differently, and our clients have no idea how or if it works, but they think it saves them money. Why would we ever ask hard questions about it, if our clients are mesmerized by the technology and never stop to ask, ‘Why do we assume that every available ad impression on even the worst website is worth being monetized if the price is low enough?’”
In the United States, an estimated $1.62 billion was spent on misinformation websites. Shifting all of that to the websites of legitimate newspapers would add nearly 50 percent to the fortunes of these hard-pressed publishers.
As the two leading demand-side platforms, Google and the Trade Desk enabled most of the ads that appeared on those websites carrying hoaxes about Covid and the US election, or Russia-Ukraine misinformation and disinformation. They could easily have eliminated them from their inventory of impressions for sale.
In fact, when Russia invaded Ukraine, Google announced, “Due to the war in Ukraine, we will pause ads containing content that exploits, dismisses, or condones the war.” However, the company limited that suspension to the two most obvious Russian propaganda sites, RT and Sputnik News, ignoring the hundreds of others that the Russians use to promote disinformation.
The Trade Desk has said that advertisers, ad agencies, and other intermediaries in the programmatic commerce chain are free to license and use exclusion or inclusion lists from any vendor, including NewsGuard, where they can be layered into the advertising buys transacted on its platform. A key barrier has been the reluctance of those involved in the programmatic chain of commerce to take responsibility and concede that their system has a major weak link.
Many involved seem unwilling to own up to the glaring reality that the “infodemic” of misinformation and disinformation, propelled by the social media platforms’ recommendation algorithms, requires a brand-safety approach that restores some degree of focus on the nature of the content the publisher is producing.
For example, after that Hertz ad appeared next to that vicious Paul Pelosi hoax, I was introduced through a mutual friend to the CEO of Hertz. When I told him about his ads appearing where it was obvious no one at Hertz would want them to appear, and explained the option of licensing exclusion or inclusion lists to apply a filter to the programmatic process, he referred us to his chief marketing officer. She referred us to Hertz’s advertising agency. After an initial conversation, they did not respond for about a month, whereupon we received a note from them assuring us that the appearance of that ad had been a onetime glitch that had been fixed. When we saw Hertz ads continuing to appear on that site and on Russian disinformation sites, the CEO told us to speak with his new head of marketing. She never responded to our follow-ups.
This pass-the-buck, it’s-just-a-glitch reaction was not unusual. Often, when we took our lists of clients we had found advertising on Russian propaganda or health care hoax sites to executives at ad agencies, they were defensive, even hostile. How were they going to tell their clients that a problem which they had assured them was already solved—with the services of the incumbent high-tech brand-safety companies that they were already having their clients pay for—had, in fact, not been solved? As the problem has become more obvious amid the crises roiling the world and producing an increasing array of brand-unsafe advertising venues, a growing number of executives and marketers who work for brands that have long paid attention to the values they project and/or the efficiency of their advertising have been more responsive when confronted with a gap in the system for which no one person is actually responsible. Still, many in the industry remain reluctant to acknowledge the flaw in what is now a nearly 20-year love affair with their dazzling technology. The system largely remains broken.
As a result, the news and information ecosystem that is so important to a functioning democracy and civil society has suffered a double whammy. First, as we have seen, the social media platforms’ recommendation engines have promoted misinformation and disinformation. Second, we have now seen how programmatic advertising has provided financial support, even from the likes of Warren Buffett, for that misinformation and disinformation, because the system is auctioning off access to the targeted person with no regard for the accompanying content.
A 2021 data analysis conducted by Comscore, a media monitoring and data company, estimated that $2.6 billion in advertising revenue was sent to publishers of misinformation and disinformation by programmatic advertisers in 2020. In the United States, an estimated $1.62 billion was spent on misinformation websites. Online advertising on all US newspapers was only about $3.5 billion in 2020, meaning that shifting all of that $1.62 billion to the websites of legitimate newspapers would add nearly 50 percent to the fortunes of these hard-pressed publishers. As mentioned, the 2023 report by the Association of National Advertisers estimated a much larger number for what it called low-quality made-for-advertising websites. Its finding that the average campaign for a big brand appeared on 44,000 websites produced a far higher estimate—that 14 percent of advertising dollars were spent on made-for-advertising sites lacking any editorial quality. If true, that would translate into more than $40 billion per year internationally. Moreover, the brand-safety companies’ “solution”—long lists of blocking words—has undermined the business of those at the other end of the content spectrum offering the most valuable news and information, by blocking the ads that they would otherwise receive.
From The Death of Truth by Steven Brill, to be published on June 4, 2024, by The Knopf Doubleday Publishing Group, a division of Penguin Random House LLC. Copyright © 2024 by Steven Brill.
If you buy something using links in our stories, we may earn a commission. This helps support our journalism. Learn more.