Back in January, YouTube surprised me by announcing its intentions to push back on "AI slop." Those aren't my words: The company directly called out AI slop on its platform, and declared a mission to reduce "the spread of low-quality, repetitive content" on YouTube. In an era where social media engagement is worth more than anything, YouTube fighting back against a "genre" that pulls in a ton of traffic (and spreads misinformation) means something.
This isn't necessarily a new venture for the company. YouTube has been applying labels to AI-generated or manipulated videos since 2024, though that came with a large caveat: The creators needed to disclose that they used AI for the upload. It was possible, therefore, to publish an AI-generated video on the platform without admitting as much to YouTube. The platform would then show that video to thousands or even millions of viewers, many of whom may never know the clip isn't actually real. Luckily, that appears to now be changing.
YouTube's new AI labels
On Wednesday, YouTube announced a new AI labeling system that affects both viewers and creators alike. First, there are the labels themselves. YouTube says it's making these AI labels far more obvious, whether you're watching the video on the app or through YouTube's website. For long-form videos (otherwise known as traditional YouTube videos), you'll see the label below the video player, above the description. It's even more obvious for Shorts: YouTube will add a label overlaid on the video itself, so you can't miss it.
The label is pretty obvious on any video you watch, however, which is excellent. AI-generated videos are only becoming more realistic, which puts viewers at greater risk of exposure to disinformation. Sure, some AI videos are obvious (at least to most), but still others look photorealistic. Without these clear AI labels, many viewers may believe things that simply aren't true. That could be as harmless as an AI bunny jumping on a trampoline, or as dangerous as a politician "admitting" to crimes they didn't actually commit.
YouTube is also rolling out new AI detection
All of this is irrelevant if YouTube doesn't know the video in question was generated with AI, of course. If the company is still running on the honor system, the labels will only be as useful as the honesty of the creators. At least, that was the case. Going forward, YouTube is supplementing its AI disclosure requirements with an automated AI detection system.
What do you think so far?
Starting this month, if YouTube's system detects a video uses AI, it will automatically apply the new AI label, even if the creator did not disclose the AI use themself. That's bad news for would-be spreaders of misinformation: In theory, you should no longer be able to sneakily upload a hyperrealistic AI video on YouTube and pass it off as a legitimate clip—assuming that YouTube's AI detection software is actually accurate.
Which leads to the immediate downside of the system: false positives. AI detectors don't have the reputation many believe they do, and even when they work well, they're prone to error. You certainly don't want people spreading AI videos like wildfire on YouTube, but you also don't want to punish creators whose videos are mistakenly flagged by bots.
YouTube says that creators have some control here. The company encourages creators who believe their videos were mistakenly labeled to update the disclosure status in YouTube Studio. However, it's still early days for the feature, so it's not clear how easy this process really is.
