TikTok Adds New Inventory Filters to Ensure Safe Ad Placement in the App

TikTok’s looking to provide more ad placement assurance for brands with a new ‘Inventory Filter’ process, that will enable advertisers to ensure that their content isn’t shown alongside potentially offensive clips in the app, if they so choose.

As explained by TikTok:

“The TikTok Inventory Filter provides advertisers with more control over the content that appears adjacent to their In-Feed Ads on the For You page. Now available in 25 countries and 15+ languages, the TikTok Inventory Filter is both a major milestone for brand safety and suitability on TikTok and a foundational technology for future advertising offerings.”

Built into the ad set up process in TikTok Ads Manager, the new Inventory Filter uses machine learning to detect variable concerns within user uploaded content, filtering each upload into one of three categories.

• Full – This category includes high risk content, which is material that doesn’t violate TikTok’s rules, but may push the boundaries of what some may find acceptable. This could include uploads which feature ‘glamorization or gratuitous depiction of mature themes’
• Standard – This tier excludes high risk, non violative content, but may still include some videos that depict mature themes
• Limited – Ads in this classification will only appear next to content that does not contain mature themes

That will give advertisers more control over where their promotions appear in the app – though the accuracy of TikTok’s automated classification will be the real test in this process, and the true measure of how valuable, and useful, these options can be.

TikTok says that, thus far, hundreds of major brands have leveraged its Inventory Filter within the testing period – “with every campaign seeing exceptional results, to the tune of 95%-99%+ “safe delivery rates”.

TikTok further notes that the technology is able to learn in real time to hone its classification, and evolve with emerging content trends.

That also points to just how advanced TikTok’s systems are at detecting content within video clips, which is one way in which it fuels its highly addictive ‘For You’ feed.

Indeed, back in 2020, TikTok came under fire after leaked internal documents revealed that the app’s moderators were being told to suppress posts created by users deemed ‘too ugly, too poor, or too disabled for the platform’.

At TikTok’s scale, there’s no way for its moderation teams to have manually checked every upload against these parameters, which suggests that TikTok’s systems are able to detect specific elements like these within uploaded clips, in order to then highlight such for moderators to review.

TikTok says that these regulations were never enforced on TikTok itself, and that these parameters were only ever applied in the Chinese version of the app (and only for a short, erroneous period). But again, the bigger point of note here is that TikTok’s systems are able to detect very specific aspects within each clip, which it can then use to show people more of the same, enhancing engagement in its feed.

In other words, TikTok can determine that a video depicts, say, a young girl with tattoos, wearing a low cut top. If you engage with that, by watching it through or liking it, it can then show you more clips which include those same, increasingly specific parameters.

That seems concerning, though it does make sense that TikTok’s algorithmic matching can be so enticing, given its capacity to determine more finite elements within each clip.

Which may actually make this new tool more accurate – but again, it does point to the advanced entity detection in TikTok’s system, which is the secret sauce of its app.

In this application, that could help to ensure safer ad placement. But in general, that capacity does seem potentially problematic.

TikTok’s new Inventory Filter is now available to advertisers in Australia, Brazil, Canada, Egypt, France, Germany, Indonesia, Italy, Japan, Kingdom of Saudi Arabia, Malaysia, Mexico, Netherlands, Philippines, Poland, Singapore, South Korea, Spain, Sweden, Thailand, Turkey, United Arab Emirates, United Kingdom, United States, Vietnam.