TikTok's AI slop slider shows what YouTube and Meta won't ship

There is a pointed line from The Verge that captures the mood around AI content moderation pretty well. In a piece headlined "Let us filter AI slop, you cowards", Jess Weatherbed argues that the major platforms have been comfortable labeling AI-generated content but conspicuously unwilling to let users actually filter it out.

The proximate cause is TikTok, which is testing a sliding setting that lets users see more or less AI-generated content in their feeds. To make that filter usable, TikTok is pairing it with invisible watermarking on AI uploads, creator self-labeling inside its Content Studio, and its own detection models. The company says it has already labeled more than 1.3 billion AI-generated videos using these techniques. That is the part worth pausing on. A slider is a much stronger consumer signal than a badge, because it changes what gets distributed, not just what gets disclosed.

By contrast, what Google's YouTube and Meta's Instagram, Facebook, and Threads currently offer is mostly disclosure. YouTube tags heavily AI-edited uploads with visible "AI-generated" labels and uses SynthID invisible watermarking. Meta puts an "Imagined with AI" badge on photorealistic images made with Meta AI and embeds C2PA and IPTC metadata. Both let users nudge their feeds away from synthetic imagery, but the framing in the reporting is blunt: neither gives a clear-cut filter. Pinterest, smaller and less central to the AI debate, already started letting users filter out some AI-generated content late last year.

The honest caveats are real. Invisible watermarks are not as robust as the press releases imply, and security researchers have pointed out that C2PA "is not designed to be hard to remove. It's just metadata," strippable via online tools or a screenshot. A filter that depends on watermarks plus creator honesty plus a detection model will let plenty of slop through, and may flag legitimate AI-assisted work along with the dross. What the reporting does not give you is how aggressively TikTok will tune the slider, what happens to creators of borderline AI-assisted content, or whether advertisers will start asking for the same filter on the buy side.

The interesting bit going forward is competitive pressure. If TikTok's slider becomes something users actively turn down, Meta and YouTube will have a harder time arguing that a corner-of-the-video badge is the same thing as user control. That is the move worth watching.