IAS and Reuters Just Put a Number on Keyword Blocking (It's 54% of Safe News)
If you run programmatic and haven't audited your brand safety keyword list since 2020, there's a decent chance you've been quietly walking away from half of your safe news inventory. Reuters just made the receipt public.
On April 6, Integral Ad Science and Reuters published the results of a test they ran across Reuters.com. IAS used its own contextual targeting layer to clear pages it deemed low- or medium-risk for brand suitability, meaning the content passed the kind of AI-driven page analysis that most of the programmatic industry now treats as the modern replacement for blocklists. Then they checked how many of those same "safe" pages would have still been blocked by a typical keyword blocklist. On Reuters' news section, 54% of them would have been flagged and demonetized anyway.
That's Reuters. One of the most buttoned-up wire services on the planet.
The test, and why it actually matters
IAS ran the study in two passes. The first hit Reuters' news content in June. The second looked at the lifestyle page in September. In both, the methodology was the same: let IAS's Context Control Targeting layer decide which pages were brand-safe, then cross-reference against a standard keyword block list.
On news, 54% of the pages IAS cleared would have tripped the blocklist. The lifestyle page was a different shape. The overlap was smaller in page count, but the URLs IAS identified as safe and that a standard blocklist would still have blocked accounted for 13.5% of that page's ad impressions during the test window. For a publisher with millions of monthly impressions, 13.5% of lifestyle impressions is a measurable tranche of inventory being excluded for words that probably no longer mean what the blocklist assumed they did.
Take "shooting" as the obvious example. It sits on most legacy keyword lists and was added for good reason. It also blocks every story about NBA shooting percentages, vaccine shots, photo shoots, and Reuters photography features. Most of those are as brand-safe as advertising gets.
The $2.8 billion number that never got fixed
This problem isn't new. It's just that most teams have stopped looking.
A 2019 report from CHEQ and the Merrick School of Business at the University of Baltimore calculated that crude keyword blocking cost U.S. news publishers $2.8 billion in ad revenue that year alone. The study estimated that 57% of brand-safe content was being incorrectly blocked by blacklists, even though roughly 40% of global premium inventory qualifies as neutral or brand-suitable.
Six years later, Reuters' number is in the same ballpark. Which is either a comforting confirmation of a stable benchmark, or more honestly, evidence that nothing material has changed in the meantime.
Stagwell has been pushing the same argument for years: advertising against news is good business, and the fearmongering around it is draining journalism for no measurable lift in brand safety outcomes. You don't have to buy Stagwell's full thesis to see that running a 2020-era block list against 2026 news is a weird way to spend money. "Pandemic." "Quarantine." "Variant." None of those words earn their keep on a keyword list today the way they did four years ago, and most of us never went back to delete them.
It's technical debt with a budget line attached.
Your list is probably worse than Reuters'
Reuters is Reuters, and that's worth a fair caveat. Their editorial standards are strict, legal review is real, and the inventory coming off their pages is about as clean as programmatic news gets. The 54% figure is probably the floor, not the ceiling.
If you're buying through general news packages, running broad programmatic across the open exchange, or using inherited block lists from a previous agency, the waste is almost certainly higher. Adalytics has documented tens of thousands of individual news stories being labeled unsafe by brand safety vendors that, on inspection, had nothing objectionable in them. The issue is structural.
The uncomfortable part is that the block list is usually invisible. Most clients never see the keyword file. Agencies rarely pull it for review. DSPs don't surface "you would have bought this impression but one of your keywords matched" as a line item. So the waste compounds silently, and nobody on the team has the phrase "blocklist audit" on their OKRs.
On paper, this sounds like an upgrade problem you can solve by buying new brand safety tech. Sometimes it is. But you can usually get most of the value for the price of an afternoon.
Pull the list. Read the first fifty entries out loud.
If you want one thing to do this week: ask your media team or trading desk for the current brand safety keyword block list you're running. The actual file. Not a summary, not a dashboard, not a vendor certification. The list itself.
Then read the first fifty entries out loud.
For each one, ask whether the word still means what your 2020 block list assumed it meant. "Pandemic" made sense in April 2020. It's mostly fossil in April 2026. "Violence" blocks every story about economic violence in a research report. "Shot" blocks vaccine news, NBA coverage, and photo shoots. If a specific entry is catching more safe stories than unsafe ones, it's costing you impressions and protecting almost nothing.
Set an audit cadence after that. Quarterly is fine. Annual is probably the minimum. Digiday's reporting on advertisers reconsidering blocklists notes that the teams seeing real lift are the ones moving from static exclusion to dynamic inclusion, the so-called "include what you like" model. That's a longer project. The block list audit is the one you can finish before your 10 a.m. standup.
A useful benchmark: if nobody on your team can name the last time the block list was touched, it's been too long. I'd set the bar at updated in the last 90 days. Anything older is probably costing more than it's protecting. My rough prediction, from what I've seen across industry reporting and practitioner discussions: audit a stale list and you'll typically recover somewhere between 10% and 25% of previously-blocked news inventory on the first pass. That's a figure backed by industry case studies, not a vibe.
For a related pattern on how account-level hygiene is eating keyword-level decisions across paid media, we covered the commoditization of keyword strategy under broad match earlier this week. The through-line is the same: the platforms keep shipping systems that reward teams who actually look at the inputs.
What Reuters' number actually buys you
I don't think most marketing teams will rush to fix this. Block lists are boring. Audits are unglamorous. And the people who catch the waste rarely get credit for it, because "we recovered 15% of our news inventory" is a harder brag than "we launched a new creative." That's probably why the 2019 number never moved the needle.
But the Reuters number is different. It has a publisher name on it. It has a vendor name on it. It has specific percentages attached. When the next brand safety conversation comes up in a QBR, it's a lot easier to say "Reuters and IAS showed 54%" than to argue from instinct. From what I've seen, that kind of ammunition is usually what moves a block list audit from the permanent bottom of the backlog into a sprint.
Worth the afternoon, at minimum. Probably worth a lot more.
Notice Me Senpai Editorial
