AI-Based Audio Is The Next Brand Safety Battleground, And It's Exploding On TikTok And YouTube

Memes that use video, images, or text are obsolete. Memes with voiceovers are currently trending on social media sites, particularly TikTok and YouTube. This is the style that will soon be overrun with content produced by generative AI.

According to compiled data analysis by brand safety ad-tech company Zefr, the majority of these AI-generated videos show lawmakers like Joe Biden and Donald Trump spreading rumors and conspiracies. Recent instances include the two trash-talking each other while playing video games, Trump declaring “America is tyranny and fascism,” and Biden praising the fact that “Epstein didn’t kill himself.”

According to Zefr study, the volume of AI-generated videos featuring politicians on social media as of early 2023 has already surpassed that of all videos from 2022 by more than 130%.

Or Levi, head of data science at Zefr, claims that these videos have received more than 2 million likes and over 11 million views. Additionally, over 90% of these videos can be made money, so businesses are presently running advertisements next to the content. The business refused to disclose how much was spent on advertisements that appeared next to AI-generated audio misinformation.

According to Kieley Taylor, global head of partnerships at GroupM, “there is a genuine risk around deep fakes and [AI-generated audio] technology evolving in time for the next election cycle.

Recent developments in AI have made it simple for dishonest people to spread false information on social media. These developments are coupled with the availability of text-to-speech models like Microsoft’s Vall-E or open-source tools like fakeyou.com. Despite social media firms spending money on content moderation, there is concurrently a growing concern among advertisers to combat misinformation. However, existing misinformation tools like static inclusion lists and keyword blocklists might not be sufficient.

Additionally, ChatGPT was discovered to be used by creators to write scripts for propaganda films. Despite stringent content guidelines, text-to-image tools like Vall-E have been used to depict politicians as reptilian or lizard people members of the Illuminati.

AI solutions outside of social

Advertisers can use keyword blocklists or AI-driven contextual tools from third-party verification firms like DoubleVerify and Integral Ad Science (IAS) to filter ads that are brand safe and brand suitable outside of TikTok and Youtube, where these videos are presently trending.

Both are used by GroupM’s brand partners, according to Taylor, who reserved block lists for objectionable or illegal material. Depending on the client’s risk tolerance, contextual AI tools can help refine possible content adjacencies on quality publishers who may have a range of content themes varying from appropriate to inappropriate.

Additionally, the sector is filtering out subpar suppliers by being more selective about the business partners it works with.

Deva Bronson, EVP, global head of brand assurance, Dentsu Media, stated that any attempts to increase supply path transparency will let clients know precisely where their advertisements will appear.

However, Andrew Serby, chief commercial officer at Zefr, notes that some tools, such as keyword blocklists or static inclusion lists, may be useless on social platforms, which require content analysis outside of website domains.

In order to measure misinformation across social media, advertisers must therefore look at more scalable AI fact-checking solutions (Zefr is one company that offers such tools based on training data shared by fact-checking organizations, but ad buyers interviewed for this article weren’t currently using them).

Fears extend across the audio landscape

Audio businesses like iHeartMedia are creating new tools to stop brand safety concerns from transferring from social platforms to podcasts. According to the Interactive Advertising Bureau, the (relatively small) $2 billion podcast ad industry is expected to continue growing.

According to Taylor, podcasting is expanding in terms of listenership, audiences, and community monetization.

But as Bronson points out, partners like NPR or Spotify, whose best interests lay in the content moderation of those podcasts, typically offer monetizable podcasts.

She stated, “I haven’t had any worries come across my desk about misinformation being rife in podcasts that our clients are presently purchasing.

AI-Based Audio Is The Next Brand Safety Battleground, And It’s Exploding On TikTok And YouTube

Memes that use video, images, or text are obsolete. Memes with voiceovers are currently trending on social media sites, particularly TikTok and YouTube. This is the style that will soon be overrun with content produced by generative AI.

Download The Radiant App To Start Watching!