No brand marketers—or at least very few—want their ads showing up alongside content that promotes lynching Black people, spews racial slurs, delves into the virtues of horse genitalia, or shows a popular video game character masturbating.

Yet hundreds of brands have been doing just that in recent months, albeit likely in front of small audiences, despite investments of millions of dollars in brand safety controls.

A report out today from research firm Adalytics finds ads from hundreds of brands and every major agency holding company being served on a variety of objectionable content from users of fan wiki site Fandom. The site contains a variety of generally unobjectionable content, such as the extensive Memory Alpha wiki covering historical details of the “Star Trek” franchise. But it also has less savory content, which a spokesperson said amounts to less than 0.08% of its 50 million pages.

Fandom also is a publisher client of DoubleVerify and Integral Ad Science, whose tags appear in source code of the ads in the Adalytics report, in several cases independently verified by Ad Age. Those tags, however, were generally for publisher services, and it’s unclear from the code what brand safety settings advertisers were using or whether Fandom had any controls in place to automatically prevent objectionable pages from triggering programmatic bids.

One marketer contacted by Ad Age only learned after seeing the report that by putting Fandom on an inclusion list it overrode any specific keyword or content avoidance restrictions offered by the company’s verification vendor. That means the only option for avoiding objectionable content is to block the site as a whole. For this marketer, the Adalytics report raised questions about whether the industry’s entire effort to ensure brand safety makes sense or amounts to any more than empty posturing.

Fandom itself is only one outpost with nine-figure revenue in a vast digital marketplace. But the apparent failure to block ads from appearing on content that should be easily spotted using keyword blocking on the open web raises questions of whether verification vendors can prevent brand safety issues in more complicated settings, such as giant social media platforms and video.

Despite glowing talk from measurement vendors about applying advancements in artificial intelligence and machine learning, it may be technically and economically infeasible at this point to provide anything close to “zero tolerance” of brand safety risks in digital media, according to some executives who saw an advance copy of the report.

Racial slurs and horse genitalia

Keyword blocking for brand safety has been an established practice in the measurement industry for many years, which would make it seemingly impossible to find ads from hundreds of major brands showing up on a page where the N-word is repeated more than 600 times, or on another with a video that advocated for the murder of Black people.

Yet Adalytics found exactly that, and Ad Age was able to independently verify many such ad placements, along with the presence of verification vendor tags in the related source code.

Prior to either Fandom or one of its moderators taking down comments and labeling the pages as candidates for deletion, the N-word page and the lynching page were serving up ads for numerous brands. Even after that, a page labeled “No Cock Like Horse Cock,” and another titled “Super Mario 3 Masturbation,” the latter showing a video of the popular video game character gratifying himself amid his adventures, continued to generate ad impressions from marketers. The latter included ads from Super Mario’s gaming platform Nintendo. Nintendo did not respond to a request for comment by deadline.

Vendor and Fandom comments

A spokesperson for IAS noted the company had not been able to see the full report in advance (Ad Age provided excerpts for comment).

“We’re committed to helping our customers safeguard and scale their brands, and we remain focused on delivering the industry’s most trusted and transparent measurement and optimization offerings by partnering with industry bodies and working alongside our advertiser and publisher customers,” according to an IAS statement. “IAS is MRC accredited for property-level brand safety ad verification on the open web for desktop and mobile web environments.”

“At DoubleVerify, we maintain the highest standards of accuracy in content classification,” the company said in a statement. “A new report by Adalytics has misrepresented our classification and demands correction.”

DV noted that having a publisher on an inclusion list would supersede content avoidance specifications. Most advertisers use pre-bid avoidance across the majority of campaigns, according to DV, but not even content identified as negative may be filtered out for various reasons. For example, one of the advertisers cited in the report chose a “monitor-only” approach for the majority of campaigns, according to DV.

“The results in this report are entirely manufactured,” DV stated, “from the omission of client campaign setup information to the methodology itself, where the researcher arbitrarily searched for racist terms.”

Fandom, in a statement, said: “We do not condone the posting of inappropriate and racially insensitive material anywhere on our UGC platform. This content is not allowed per our guidelines, and it won’t be tolerated. Ensuring brand safety on our platform is of the utmost importance to us and we take matters like this very seriously. We also welcome the efforts of any company attempting to protect the rights of our advertisers. However, this report identified inappropriate content on old, extremely low trafficked wikis, which is why it was not flagged via our current moderation systems or Google’s Ad Server, both of which monitor our active wikis. Despite the fact that these narrow examples were not a problem at scale, we have added additional safety measures to proactively turn off ads on low trafficked wikis that don’t trigger flag thresholds.”

“What this report does is identify an industry-wide problem,” according to the statement. “There is a critical need for all publishers to be vigilant and proactive in this area and to work with key vendors’ tools and resources to ensure industry standards are not only met but exceeded.”

While the Adalytics report focused on Fandom, the issue of brand ads appearing on objectionable keywords, comment posts and other pages happens elsewhere, and the company planned to issue an addendum on Wednesday showing occurrences elsewhere.

Also read: Oracle tried to sell ad business to DoubleVerify and IAS

How the study started

The Adalytics report came in response to a request from a marketer client to check whether brand safety assurances of complete brand safety from the company’s agency and verification vendor (DoubleVerify) were reliable. Ad Age was able to confirm this with the marketer, who, like others interviewed, spoke on the condition of anonymity because they were not authorized by their companies to speak with the media.

That marketer, like three others who are clients of DV or Integral Ad Science, expressed surprise and outrage that the firms didn’t screen out even the most egregious content such as racial slurs using technology as simple as keyword blocking.

The four marketers interviewed by Ad Age said their companies spend between $2 million and $6 million annually with verification vendors, in part to prevent such placements against objectionable digital content. All of the marketers interviewed said they believed they had instructed verification vendors to use stringent brand safety controls. One marketer said their company had even instructed a verification vendor to block ads on all user-generated content on the open web, but still had ads show up on Fandom.

“There’s a lot of misreporting of the capabilities of the verification partners,” said the first marketer after reading the Adalytics report and getting data regarding their company’s own ad placements. The marketer questioned whether verification vendors have technology that’s “fit for purpose,” and whether they’re actually monitoring every website.

“We want to ensure that what we’re paying for is what we’re getting, because we’re paying them to do things like pre-bid blocking as well as post-bid reporting,” the executive said.

Need seen for page-level exclusions

A second advertiser, who said their company had specified it wanted no ads to appear against user-generated content on the open web, said Fandom never should have been part of its buy in the first place, yet the brand ended up with several ads showing up on UGC pages with racial slurs and hate speech.

“Technology has been paid for that clearly doesn’t work,” said a third marketer after reading the report and seeing examples of their company’s ads showing up on pages with objectionable content. The report, the marketer said, indicated a clear need for URL and page-level blocking and post-campaign reporting that the company isn’t getting now.

“There is now a market opening for a new vendor who can address this and solve it,” said the marketer, who added that Adalytics may fill that opening. But Adalytics does not currently compete in the brand safety monitoring, and CEO Krzysztof Franaszek in a statement said the company has no current plans to do so. "To crawl, render, and accurately classify internet-scale, full-page URL level data with multimedia, some of which is gated behind user logins, in real time would likely require billions of dollars in computing resources," he said. "As such Adalytics does not currently envision offering such a commercial capability."

‘What are we even doing in the industry?’

A fourth marketer—the one who questioned whether the report exposes a broader flaw in the industry's brand safety stances—said it shows the ad industry’s quest for brand safety may have become absurd.

“We’re impacted, and that sucks,” the marketer said. “But it’s to the point that I’m questioning what are we even doing in the industry. I’m past the point of technical outrage to asking what the hell do we even mean by brand safety? Am I just chasing a poorly articulated goal that has no actual ability to be met?”

Fandom was on its inclusion list, the marketer said, but on the expectation that the company’s verification vendor would block placements on objectionable content.

“The pre-bids I make should theoretically catch a URL that literally says ‘Big Black Dildo,’” the marketer said, adding that Fandom being on the company’s inclusion list apparently overrode that control.

“No rational marketer would think that way,” the marketer said, “but that’s making it clear to me that we’re just chasing stupidity.” The executive wondered aloud whether their company should just forget about brand safety and advertise on Pornhub.

Competitive scrum

The Adalytics report comes as DV and IAS are scrambling to pick up clients left behind by the impending shutdown of another competitor, Oracle, whose advertising measurement business is expected to shutter by the end of September. Oracle in June announced it would exit the business, and decided to simply shut it down rather than sell it to a rival.

The commitments made in marketing materials are fairly expansive. DV on its website touts its Authentic Brand Suitability product as “one of the most advanced brand safety and suitability and fraud solutions in the market today.

“Our solution maximizes buying effectiveness and reduces waste by applying your post-bid brand safety and suitability and fraud controls in a pre-bid environment. Using ABS, you can create a centralized set of brand safety and suitability and fraud controls—allowing for customized protection that is easily deployed across multiple programmatic buying platforms and campaigns using a single segment ID.”

An IAS description of its brand safety offering states: “We don’t simply provide a binary brand safety solution—that can unnecessarily limit scale. Our standard brand safety content categories are powered by a robust machine learning based methodology that provides more accurate and scalable coverage across all formats and environments.”

Neither company, however, mentions that marketers can’t block individual objectionable keywords, URLs or pages on sites they otherwise have on inclusion lists.