A recent report by Adalytics is causing a stir among advertisers and brand-safety experts by questioning the effectiveness of AI systems used for brand protection. Released today, the report highlights that numerous brands have appeared next to unsafe content on user-generated content (UGC) websites like Fandom wiki pages.
Despite the presence of brand-safety technology from companies like DoubleVerify and Integral Ad Science, the report reveals significant issues with ad placements and brand safety.
The research was prompted by a request from a multinational brand’s global head of media, who sought an evaluation of its ad placements. The findings have led to new concerns about the effectiveness of AI in maintaining brand safety.
The report suggests that the technology might not be functioning as intended, raising questions about the transparency and reliability of AI systems in this context.
Advertisers and brand safety experts have expressed skepticism about the performance of AI tools. The report indicates that these tools are failing to meet the standards promised by brand-safety companies.
Issues have been identified with how these tools handle content that includes racial slurs and sexual references, which should be easily identifiable but are slipping through the cracks.
Further scrutiny reveals inconsistencies in how AI tools categorize web pages. The report shows that some pages with problematic content are labeled as low risk, while reputable news sites like The Washington Post and Reuters are flagged as medium or high risk despite not containing harmful content. This discrepancy has led to doubts about whether the technology is being marketed accurately.
The report’s findings suggest that the technology might not be as effective as advertised, leading to concerns about whether the substantial investments in brand safety technology are justified. Critics argue that the current solutions offer minimal improvement and that the industry may be delusional about the capabilities of these tools.
Jay Friedman, CEO of the Goodway Group, has criticized the technology’s effectiveness and called for greater transparency in brand safety solutions. He argues that the industry needs to provide more detailed reporting and a clearer understanding of how these tools work, given the significant amounts of money spent on them.
DoubleVerify has responded by accusing Adalytics of selectively searching for problematic terms and misrepresenting their technology’s capabilities.
They claim that the report overlooks crucial information about their pre-bid and post-bid settings and incorrectly correlates code with advertiser actions. However, some sources support Adalytics’ findings, suggesting that the report accurately reflects issues with both publisher and advertiser tags.
Fandom, which hosts a vast amount of UGC, has stated that it has not yet seen the full Adalytics report but acknowledges that the examples provided reflect a broader industry issue. They claim that their moderation systems, including multiple vendors and an internal safety team, are effective in managing inappropriate content. Despite this, they are implementing additional safety measures to address the issues highlighted in the report.
In response to the report, advertisers are calling for more granular reporting to ensure that they can assess URL-level pages rather than just domain-level reports. There is also discussion about potentially changing the economic model of brand safety, such as paying based on blocked content rather than pre-bid technology.
Meanwhile, experts question the feasibility of achieving 100% brand safety with current technology, suggesting that real-time contextual analysis of web pages remains a complex challenge.
