Snapchat is the latest platform pulling back on recommending fully AI generated content

[EventHorizon63]

Snap announced its Spotlight feed will stop pushing wholly AI generated videos to the top of the feed, favoring authentic human made content instead, which puts Snapchat squarely alongside YouTube, LinkedIn and Substack, all of whom have rolled out similar moves against so called AI slop in just the past couple of weeks

The term itself has become the go to shorthand for the flood of low effort AI generated writing, images and video that's spread across the internet as generative tools have gotten dramatically easier to use and dramatically cheaper to produce content with at scale, and this coordinated wave of platform responses suggests the major players are genuinely feeling pressure from users tired of feeds increasingly clogged with it

What makes this notable is the timing, four major platforms across genuinely different categories, short video, professional networking, newsletters and now social messaging, all making similar moves within roughly the same fortnight suggests this isn't one company reacting to its own specific backlash, it looks more like an industry wide recognition that recommendation algorithms optimizing purely for engagement had started actively rewarding cheap AI generated volume over anything a human actually put real effort into making

The actual mechanics of how each platform defines wholly AI generated versus AI assisted content is going to matter enormously here, a video with AI generated background music or a caption cleaned up by a language model is a very different thing from a video where every frame and every line of dialogue came entirely from a generative tool, and getting that distinction wrong in either direction risks either letting genuine slop through or unfairly punishing creators who used AI as a genuine tool rather than a full replacement for their own effort

This also raises an obvious enforcement question that none of these announcements have fully answered yet, detecting AI generated video and images reliably at the scale a platform like Snapchat operates at is a hard technical problem on its own, and it remains to be seen whether this ends up being a meaningful algorithmic shift or mostly a policy statement that's harder to actually enforce consistently than it is to announce

Snapchat joins other platforms in the fight against 'AI slop'www.bbc.com

[RandyOrton]

The timing across four different platforms within two weeks is the real story here, feels less like coincidence and more like everyone quietly noticing the same user backlash building at roughly the same moment

[Taker92]

Substack being on this list surprised me a little, feels like an unusual platform to have a meaningful AI slop problem compared to something purely video and image driven like Spotlight

[Zenith Theo]

Curious how much of this is an actual algorithmic change versus a PR statement timed to get ahead of criticism, plenty of platforms have announced content quality initiatives before that quietly changed very little in practice

[RandyOrton]

Engagement optimized algorithms rewarding cheap AI volume over genuine human effort is exactly the underlying problem here, this was always going to happen once the cost of producing content dropped near zero while the algorithm still only cared about raw engagement signals

[Static Estuary]

Detection at scale is going to be the actual hard part regardless of how good the policy sounds on paper, plenty of AI generated video already looks convincingly real enough to fool casual human review, let alone an automated system

[Rapid Andrew]

The wholly AI generated versus AI assisted distinction is going to be a genuine mess to police fairly, creators who use AI tools as part of a real creative process shouldn't get lumped in with pure slop farms just because a generative tool touched the project somewhere

[CipherMonk37]

Authentic human made content becoming a marketing term platforms actively promote is such a strange sign of where we've ended up, feels like something that should have just been the assumed baseline a few years ago rather than a feature platforms now have to explicitly call out