Editor’s note: this is a living document. We encourage everyone to contribute ideas and corrections to this article and we’ll update it accordingly. Contact support@beeler.tech with your input.

“Why does Ad Tech want to silence Killer Mike?”

That thought kept rolling around my head while I listened to Grouchy Greg Watkins, Founder of AllHipHop.com, speak on a panel about his brand safety issues.

If you’re not hip to hip-hop, Killer Mike is part of the rap group Run the Jewels and, despite his name, is a very positive force in his community as well as overall American political discourse. His name, of course, is not brand safe, and a publisher like AllHipHop.com producing content about Killer Mike will find many advertisers avoiding those pages.

Since that panel, I’ve been trying to think through how to move the brand safety conversation forward. I spoke with the Beeler.Tech publisher community and partners about ways to make this less painful for all involved.

I narrowed in on keyword blocklists because if you’ve never reviewed a blocklist, it’s a fascinating read. They are typically Excel spreadsheets listing hundreds if not thousands of keywords and phrases that the advertiser doesn’t want their advertising adjacent to. They’re the reason advertisers are unlikely to want to appear alongside content about Killer Mike.

But in researching keyword blocklists, I found out we can’t get rid of them. Keyword blocklists are one aspect of a bigger conversation about brand safety, but believe it or not, they have a place. My guess is most people across the industry—from brands to publishers—don’t really understand how they actually should be used.

So instead, let’s talk about why keyword blocklists exist, why they are needed, and how to make them better.

What we all get wrong about keyword blocklists

Keyword blocklists are focused on the words in the URL string of a webpage. Verification services are not looking for the word “rifle” in the article and blocking ads based on its existence in the context of the article. “Rifle” would have to be in the URL of the article for a real-time block to occur from a keyword blocklist. That doesn’t mean a different verification tool wouldn’t flag the word as part of a larger content analysis looking for instances of violence. The word as it is, doesn’t flag anything on its own. Keywords blocklists are only focused on the URL string.

My “rifle” example is still problematic. One could write a headline about how someone rifled a baseball from third to first base. “Round” is an even better example as every March I’m sure it appears in the URLs of NCAA basketball coverage.

More sophisticated solutions can differentiate between a rifle used in an assault and in sports. Verification services determine if an article is about firearms or a crime, and keyword-blocking words like “rifle”, “round”, “shot” and the like don’t need to be in a keyword blocklist. Categories use context, tone, etc., in addition to the actual words on the page to classify content, and are more accurate than keyword lists.  All these words in keyword blocklists are the exceptions, not the rule since the ads would be blocked anyway based on the evaluation of the content of the page.

So why create keyword blocklists at all? Why not kill them off?

Because their function is a last-resort measure to keep up with sudden changes in the news cycle. Think of it as a quick fix while a deeper analysis of the brand safety/suitability of a piece of content can be performed.

Take the bombing at the Ariana Grande concert in Manchester in 2017. As events unfolded, a keyword blocklist with “Ariana Grande,” “Manchester,” “terrorist,” and “bombing” would help prevent an ad from appearing on a related news article. Arguably this is a failure to support journalism, but the brand gets to determine if their message belongs on that page through the quick fix of a keyword blocklist.

But the news cycle changes. I’m sure not every brand considers Ariana Grande to be brand safe, but you wouldn’t want to keep her name blocked—especially when she went back to do a benefit concert for the victims. Good news is rare – shouldn’t you capitalize on it?

Keyword blocklists therefore are a stop-gap solution focused on URLs and nothing more. According to DoubleVerify, keyword blocklists fuel a very small percentage of the actual blocking going on.

So does this mean we should just let them be, inefficient as they are? I say no. If we need them, make them work better, and use the process of making them work better to open up more important discussions we need about brand safety and suitability.

Making keyword blocklists better

Some thoughts on where to start:

Brands/Agencies:

• If I haven’t been clear enough when and where keyword blocklists work and don’t work, educate yourself. Please – it does more harm than good to do this poorly.
• If you have keywords in your lists with broad goals like avoiding violent content, it’s likely that a content category would be a better tool to use. There’s no need to put ‘murder’ on a keyword list if you’re already blocking a corresponding high-risk violence or death category.
• My guess is no one is going to get a promotion for a well-groomed keyword blocklist, but they have to be maintained. Otherwise, they actually don’t serve a function and only serve to block you from reaching the audience you’re trying to reach.
• Overgeneralization of terms is blocking you from quality content. Work with your vendor on what the right level of specificity is needed.
• Make creative that works in news-related environments so you don’t have to block or you can block less and support journalism.
• Be on the right side of history and think about what your blocklist says about your brand. If DEI is important to you, are you looking through that lens when you create your keyword blocklists? If your keyword blocklist were published, would you be ashamed?
• Keep up with the younger generation and under-represented groups and how they use words differently than you do. “Slay” is a good thing. Lil Uzi Vert is a pretty good rapper. You’re missing out on entire audiences if you don’t keep up with their vernacular.
• Do your research. Example: an IAB study found that ~90% of consumers have positive or neutral feelings about brands that advertise alongside news, and ~50% of consumers believe brands who advertise in news are more relevant, more innovative, and more trustworthy.
• Inclusion lists which most verification vendors provide! If you trust the New York Times, this is how you show it and reach the audience you are looking for.

Ad Tech Vendors:

• Give publishers visibility into how their inventory is viewed, especially for inventory available on the open marketplace. Direct sold campaigns allow for there to be some discussion between buyers and sellers and adjustments can be made. For programmatic-traded impressions, no such visibility occurs. Pubs will gravitate to companies that allow them to claw back impressions from blocking.

Verification Providers:

• More visibility for more publishers beyond your immediate client base. I think there is an opportunity to distinguish yourselves by providing more tools for all publishers. Example: what if you publish an aggregated top keyword blocklist for all pubs to see – especially as they relate to recent news events.

Publishers:

• If your efforts of analyzing keyword blocklists extend beyond looking at the URL, it’s time to change your process. Honestly, if I had the scripting chops, I’d write something that analyzed keywords vs URLs so I could stay on top of potential issues.
• This is me being a troublemaker, but I would create a “block blocklist”. These would be terms that you won’t allow a brand to block. If a brand were to block words like black, gay or a religion (even the word “religion” is in keyword blocklists), I would push back. Brands are brands but publishers are brands as well and some publishers don’t see the word “gay” as problematic in any context.
• Perhaps a little bit less troublesome: publishers should have a list of words they commonly use that shouldn’t be misinterpreted. Golfers shoot and they play rounds. Provide your advertisers this list early in the conversation so they can be removed from blocklists.
• Encourage brands to invest in news-related content. Help them develop creative that work best on your pages and the content you serve. I think you might even call out in your ad slots “this advertiser supports journalism” on pages that others are blocking.
• I think we should give awards to advertisers who support journalism on tough topics.
• Push for your advertisers to use inclusion lists!

As I mentioned at the beginning of this article, I’m open to additional suggestions and modifying the content with corrections and additional viewpoints. Contact support@beeler.tech to share your thoughts.

Let’s not forget the issue bigger than keyword blocklists

Hopefully the information about keyword blocklists along with the recommendations reduce the amount of effort/time lost on a small piece of the puzzle. Since keyword blocklists are focused on just URLs, what about the technologies being used to evaluate the content itself?

Brands have a right to buy whatever impressions they want based on whatever criteria they deem important. Full stop.

Publishers have the right to publish whatever content they think will attract an audience and have it monetized by advertising from advertisers that want to be seen by that audience. Full stop.

What is lost is that when evaluating impressions is done poorly, it can punish publishers who have invested time, money, and effort into creating content. This is especially true for publishers of hard-hitting news and from people of underrepresented groups. Therefore it is imperative that we continue to up our game on all sides of the equation.

At risk is the sustainability of journalism. Misinformation doesn’t need advertising to achieve its goals and be funded. I feel as many brands err on the side of caution – let’s not let any possibility of our advertising be juxtaposed against anything that might be unsafe. What should be the first step is to be more inclusive of the sites you deem overall brand safe and stop chasing audiences into the nether regions of the internet. Those cheap impressions you buy are rotten fruit. Work with premium publishers as you define them and instead err on the side of quality audiences understanding your brand belongs right where it is served.

Nothing is more brand safe than helping educate our society. Let’s move in that direction.

Resources/Related Articles

• AdExchanger: How Brands Can Learn To Stop Worrying About Brand Safety And Love The News
• AdAge: Black media is being hurt by agency blocklists – how Group Black and DoubleVerify are looking to fix it (Press release) Editor’s note: While this highlights blocklists as the problem, the results achieved were due to adjusting other settings.
• The Drum: Blocklists are still failing advertisers and minority media
• The Drum: Tackling LGBT+ blocklists: publishers and agencies rally against digital censorship
• DV: Brand Suitability Series Part 2: Demystifying DV’s Keyword Blocklist Service
• DV: A Guide to Brand Suitability and Keyword Best Practices (download with PII)
• DV: Demystifying the Scope and Intent of Keyword Blocking
• Brand Safety Institute Knowledge Centers
• GARM Adjacency Standards Framework PDF

Appendix

A quick sampling from actual real keyword blocklists sent to publishers as of July 2023:

Arrest
Assassin
Assault rifle
Asscrack
Auto accident

Okay—if I’m a consumer-facing brand, those topics might not be great for my “put a smile on your face” promotion. (Ignore the bigger implications about a failure to support journalism for a minute—we’ll come back to that.) Instead let’s look at a few lists and cherry-pick some terms:

Metro
Round
Subway
Nudey
Boob
Booobs
Boooobs
Booooobs
Booooooobs

According to this logic, “boobs” and “boobs” with 6 “o”s are okay, but not 5 or 7. Good to know. However, for me the most fascinating one of all is (this is real):

“Death of a Salesman”

Somehow this title made a list of 1,100 keywords. I just want to meet the person who thought the tragic fall of Willy Loman needed to be included.

I have heard people mention keyword lists that include “black”. I haven’t found an example of that yet. I have found the term “ghetto black” in one list. Lots of sexual terms make the lists but nothing specifically on sexual orientation like “gay.” I did find “Jew,” “Jewish,” “Christian,” and “Muslim” on lists.