ChatGPT
Perplexity
Claude
Gemini
Pinterest now ships a setting most platforms didn't need to build a year ago: a "tuner" that lets you dial the amount of AI content in your feed up or down, rolled out first on Android and desktop before expanding to categories like art, fashion, and home décor. TikTok is testing something similar for its For You feed. Deezer just labels the problem outright rather than filtering it: the platform says 60,000 fully AI-generated tracks, more than 39 percent of everything uploaded daily, arrive on its service every day, and it detected more than 13.4 million of them last year alone. Henry Ajder, who has studied synthetic media since 2018, compared the situation to industrial-era smog: "It's going to be very, very hard for people to avoid inhaling, in this analogy."
The volume behind that comparison is real and it's not really about any individual creator's work. TikTok alone has labeled 1.3 billion clips on its platform as AI-generated. That's a scale problem, not a taste problem, the same kind of overwhelm that made spam filters and comment moderation necessary tools rather than optional ones. Platforms building dials and tuners are responding to that volume the way any system responds to noise crossing a threshold: not by asking whether any single piece of content is good, but by giving people a blunt instrument to turn the whole category down.
A newer, sharper response is showing up alongside the dials: platforms built to exclude AI content entirely rather than just turn its volume down. Cara, a portfolio site for artists, bans AI-generated work outright. Pixelfed, an ad-free alternative to Instagram, lets communities run art servers that prohibit it. A new platform called Spread markets itself around the promise of letting people "escape the flood of AI slop." diVine, the revival of Jack Dorsey's old Vine app, is building its entire pitch around "No AI Slop" and multiple detection layers to enforce it. These aren't dials. They're walls, and the fact that walls are the product now tells you how far the dial approach already feels insufficient to some audiences.
That's the part that should concern a working creator who uses AI as part of a deliberate process rather than as a spam machine, because none of these tools, dials or walls, are built to tell the difference. A feed-ranking system and a platform ban both start from the same technical signal: was this generated. They don't have a column for "generated with fifteen careful passes and real creative decisions behind it" versus "generated once and posted a thousand times under a fake artist name." The actual risk for a serious AI creator right now is getting swept into the same bucket as everything the smog metaphor describes, invisible by association rather than by quality, not having the work rejected on its merits.
None of this points toward AI creators needing a literal separate internet, walled off the way Cara or diVine wall off AI entirely. What it points toward is something narrower and more useful: spaces and workflows that sort by intent and craft rather than by medium, the same instinct driving Cara's ban just aimed the other direction, toward making deliberate work legible as deliberate instead of leaving it to get lost in the same undifferentiated pile as the slop everyone's trying to filter out. The internet doesn't need to split into an AI half and a human half. It needs better ways of showing which side of that fifteen-pass line a piece of work actually came from, and right now, that's still mostly on the creator to prove.