Something shifted this summer. Google, LinkedIn, Meta, Reddit, Snapchat, TikTok and YouTube have all introduced measures to limit, label or demote content made entirely by AI. Seven of the largest platforms on the internet, arriving at broadly the same position within months of each other.
They didn't coordinate. They responded to the same pressure: feeds filling with content that nobody in particular made and nobody in particular wanted.
Three different tools
The approaches sort into three categories, and the distinction determines how much a creator should actually care.
Demotion is the harshest. Snapchat stopped recommending videos created entirely with AI on Spotlight. Content enhanced using Snapchat's own AI tools stays eligible if it's clearly labelled, but wholly AI-generated video is out of recommendations. On a discovery surface, that's close to invisibility — you can still post, and your existing followers can still see it, but nobody new will.
Disclosure is the middle path. TikTok requires labels on AI-generated images, video and audio depicting realistic people or scenes, including in advertising and branded content. Meta automatically labels AI-generated or AI-edited advertising on Facebook and Instagram, and failing to disclose can mean ad rejection or account penalties. YouTube moved its AI disclosure label to a more prominent position — directly below the player on long-form video, and as an on-video overlay on Shorts.
Community signals are the newest. LinkedIn added a "seems like AI slop" report button on July 30 and deployed classifiers that reduce flagged posts in recommendations, notifying authors privately when a post reads as inauthentic.
The line is human input, not AI use
Every one of these policies draws the same distinction, and it's the one creators keep getting wrong. None of them prohibit AI tools. What they penalise is content with little or no original creative input.
YouTube's version is the clearest: it enforces an "inauthentic content" policy by removing monetisation from mass-produced AI content with minimal original input, while continuing to permit AI-assisted work. Snapchat's version is functionally identical, expressed through recommendations rather than monetisation.
Read together, these amount to a shared position: AI can help make something, but it can't be the whole thing if that thing wants distribution.
What changes for creators
The pipelines that broke this year were the volume plays — generate a hundred videos, publish them all, let the algorithm sort it out. Every one of the mechanisms above is designed to catch exactly that pattern, and they now operate on the surfaces where discovery actually happens.
The workflows that survive look different. AI in the middle of the process — drafting, editing, captioning, upscaling, translating — remains fine everywhere, and platforms are actively shipping tools to encourage it. Labelling as a habit rather than a reaction is now the safer default, particularly on TikTok and in any advertising context, where non-disclosure carries direct account consequences.
The unresolved part
Detection is imperfect and enforcement is inconsistent, which means some fully synthetic content still gets through while some human work with heavy AI assistance gets caught. That asymmetry will produce complaints in both directions for a while yet.
But the direction is settled. Two years ago the platform message was that AI tools were a creative opportunity. The message now is that AI tools are fine and AI output alone is not. That's a meaningful reversal, and it happened faster than most of the industry expected.
Image: Pixabay, via Pexels
