It’s no secret that YouTube is full of AI slop. The company has shown signs of trying to rein in the flood of cheap AI images, and now it’s showing it’s a little more serious about enforcing its own (still relatively lenient) standards for videos created using the technology.
On Wednesday, the Google-owned platform announced changes to the way it begins labeling AI-generated content. The most notable update is that labels are now more prominent. From now on, traditional YouTube videos will have an “AI” label directly below the video player instead of hidden in the video description. And in its vertical videos, TikTok’s imitation shorts (a format that has proven most susceptible to being hijacked by low-effort productions, especially the kind that purports to be educational), the AI label appears as an overlay at the bottom of the video.
But YouTube’s sloppy stealth marketing tactics are getting even more serious. Starting this month, the company will begin scanning for signs of “photorealistic” AI use so it can automatically label suspicious videos. Previously, AI labels were only applied if the uploader chose to make them public.
YouTube has not disclosed what systems it uses to detect AI content, but like other companies, it has adopted C2PA, a standard for embedding provenance data in AI content, and SynthID, a Google-developed tool for watermarking AI content.
The video giant said the disclosure label covers “photographic and meaningful AI-altered or generated content.” Clearly, the company understands the risks of having a ton of misleading and authentic-looking AI fakes on its platform, but it’s not going to slam all forms of it.
Unrealistic AI content (such as pointless animated slop videos targeted at children) will never receive the new front-and-center label and will continue to use the old label hidden in an expanded description. Even more fragmented uses of AI are not prominently labeled.
In any case, aside from the potential stigma that comes with the big name “AI” label, YouTube isn’t trying to penalize AI-heavy content. The changes are “designed to balance transparency and creator control,” according to the announcement, but “will not change how videos are recommended or whether they are eligible for monetization.”
Learn more about AI: Record label claims bizarre AI-generated ‘Viking rapper’ with millions of views is real person
