It’s been more than 6 months since YouTube’s #adpocalypse went into effect and there are still more questions than answers. “Why did this video get flagged?” is one we see often. Only YouTube knows exactly how their algorithm works, but we did a bit of digging to better understand not only this question, but also how it impacts your long-term recovery, which you can read over on VideoInk.
It’s not just metadata
When you first upload a video to YouTube, the title, description, and tags will be scanned for ‘risky’ keywords. The list is not known to the public, but based on our research, swearing, violence, and sexual terms will heavily increase your chances at getting dinged.
But it’s not just written words that YouTube is looking for; it’s spoken, as well.
Whether you provide them or not, captions are auto-generated shortly after upload, and while they’re not always perfect, the system is intelligent enough to decipher most of the words you verbalize – and those are also used to determine whether or not your video is brand safe.
It would be nearly impossible for an actual human to watch every video that is uploaded to the platform, so algorithms are put to use – including Google’s fairly new Vision APIs – which almost literally scans and watches your videos.
These APIs are used to not only analyze the thumbnails of your videos, but also the content within them. Based on public statements and the research that we’ve done, it appears that this system is limited right now, but currently scans for images related to violence and terrorism.
Don’t forget to appeal
If one of your videos is marked as ‘not suitable for all advertisers’, and you feel otherwise, make sure to file an appeal once it hits 1,000 views. YouTube is using successful appeals to further train their algorithms.
Recovery is possible
Most creators have seen only a few videos become demonetized, but there are a handful who have been hit hard. We’re happy to report that it’s not a sign to pack it up, but a sign to make some changes.