The brand safety industrial complex began with a reasonable hypothesis: that advertising adjacent to difficult news content — crime, conflict, pandemic, political crisis — carried reputational risk. Brands didn't want to be associated with suffering. Agencies didn't want to explain to clients why their display ad appeared next to a mass shooting. The logic was intuitive, if never rigorously tested.
So the industry built an avoidance architecture. Keyword blocklists, third-party verification platforms, brand safety floors maintained by GARM, and exclusion categories covering everything from "military conflict" to "death" to "terrorism." By 2019, the costs of this architecture had been quantified: US publishers lost an estimated $2.8 billion annually in programmatic ad revenue due to incorrect keyword blocking, according to a study by cybersecurity firm CHEQ. One in four potential ad dollars never materialized because an algorithm saw the word "shot" in a tennis article and decided the placement was unsafe.
The avoidance infrastructure grew regardless. DoubleVerify and Integral Ad Science built nine-figure businesses around the premise that brand safety required active, continuous management. The logic calcified into practice: hard news was not just undesirable as an advertising environment — it was actively dangerous.
Then researchers decided to test the premise.
In 2020, economists Andrey Simonov, Tommaso Valletti, and Andre Veiga ran a randomized eye-tracking study with 1,013 UK and US participants recruited via Panelbase. The setup: participants read nine news articles — a mix of hard and soft news from The Guardian, Daily Mail, The New York Times, and USA Today — with eight display ads embedded across the articles. One slot was left blank as a control. The researchers used Lumen Research's non-intrusive eye-tracking technology to measure exactly how long each participant's eyes dwelled on editorial content versus adjacent advertising.
The findings, published in the Journal of Marketing Research (Vol. 62, No. 2, pp. 294–315, 2025), overturned the industry's foundational assumption about how attention flows between content and advertising.
The core result was counterintuitive. Readers spent less total time on hard news articles — COVID-19 pandemic coverage and Black Lives Matter reporting — than on soft news. This made intuitive sense: people were tired, stressed, seeking escape. But advertising adjacent to hard news was significantly more effective, not less. At the mean ad dwell time of 2.76 seconds, soft news placements increased purchase probability by 1.38 percentage points. Hard news placements increased it by 2.04 percentage points. An 18 to 43 percent uplift depending on the comparison.
The average article, taking 106 seconds to read, created 0.848 seconds of additional ad attention. A 28.2 percent increase in ad attention generated by editorial engagement, not by anything the advertiser did.
The mechanism, as the researchers modeled it, is attention spillover. When readers encounter content that fully absorbs them, their eye movements become more active, their gaze more fluid. This heightened attentional state doesn't stop at the article's edge — it carries into adjacent visual space. Hard news, despite commanding less total time, apparently creates more intense per-unit-moment engagement. The absorbed reading state — the kind that produces spillover — occurs more readily when content is genuinely compelling rather than casually browsed.
The CHEQ 2019 study remains the most cited quantification of what brand safety avoidance costs publishers. Applying an observed 57 percent incorrect blocking rate to the portion of online news inventory verified as brand safe, the researchers estimated $2.8 billion in annual programmatic revenue lost to over-blocking. One in four dollars that should have transacted didn't, because an algorithm couldn't distinguish between a Reuters investigation and a Reuters fashion piece.
The examples accumulated. An article about "The Devil Wears Prada 2" film production was blocked because the URL contained "shot." Coverage of the band Arcade Fire triggered fire-related blocks. Paris remained on many blocklists three years after the Notre Dame fire. Taylor Swift's Time cover story — a major fashion feature — was flagged because it mentioned "war on feminism." Only nine of the top 100 New York Times articles in 2019 met typical keyword-based brand safety criteria.
The Reuters and Integral Ad Science study in 2023 tested typical blocklists against actual Reuters content. The finding: 54 percent of brand-safe Reuters news URLs would have been blocked by conventional keyword systems. On Reuters lifestyle pages specifically, blocked-but-brand-suitable URLs accounted for 13.5 percent of ad impressions during the test period.
DoubleVerify's 2023 revenue was $572.5 million. Integral Ad Science's was $474.4 million. These businesses were built on the premise that proximity to hard news was a risk requiring active management. The premise, as tested by Simonov and colleagues, appears to be wrong.
Some advertisers and agencies have begun acting on the evidence. The Stagwell Group replicated the finding in 2024 with a 50,000-person US survey showing that ads adjacent to Trump, Biden, crime, and inflation content performed on par with sports and entertainment across purchase intent and brand metrics. A UK follow-up with 22,116 participants found the same result. Stagwell's conclusion: current brand safety practices are too blunt, and news audiences distinguish editorial from advertising in ways that blocklists cannot.
The contrarian move is to buy hard news inventory deliberately. Not as a brand safety exception — as a performance strategy. The data suggests the opposite of what the avoidance architecture assumed: that serious journalism generates a contextual premium, not a discount.
This creates an interesting strategic paradox. The advertisers most likely to act on Simonov's findings are precisely those most committed to avoiding controversy — which is to say, the ones most likely to have been using blocklists most aggressively. The reversal requires admitting that a decade of avoidance was not just ineffective but actively counterproductive. That the $2.8 billion annual transfer from publishers to avoidance infrastructure was spent solving a problem that didn't exist.
The brands that figure this out first get to advertise next to journalism that commands engaged readers, at inventory prices suppressed by years of systematic de-monetization. The dividend is already there. It just hasn't been collected.