Platforms are fighting AI spam—which they themselves created

Snapchat has stopped recommending fully AI-generated videos in its Spotlight feed. Over the past two weeks, YouTube, LinkedIn, and Substack have announced similar measures. Platforms that have built generative tools into their products are now grappling with the direct consequence of doing so—“AI slop,” or low-quality content that is flooding their feeds.

Snap isn’t banning AI content entirely. The platform offers its own editing tools, so content that has been “enhanced or edited” using AI will still appear in recommendations. The distinction—between “fully generated” and “human-created with AI”—sounds logical, but in practice, it’s nearly impossible to apply consistently.

LinkedIn went a step further and removed its own prompt that offered users the option to “enhance” a post using AI. Instead, it brought back a simple spell-check tool. At the same time, the platform added a button to report AI-generated posts and comments. According to Product Director Hari Srinivasan, LinkedIn has blocked “billions” of attempts to post automated comments and catches hundreds of thousands of new ones every day.

YouTube has updated its monetization policies: content that is formulaic, repetitive, or generated according to a single pattern is ineligible for revenue. An earlier study uncovered dozens of channels consisting entirely of AI-generated content—some of which had millions of subscribers and generated millions of dollars in revenue.

The numbers confirm the scale of the problem. According to a study cited by Substack CEO Chris Best, up to 40% of text on social media is already generated or written to distort reality. Another survey shows that the more fake content users see, the less they trust any content in their feed—including the real stuff.

The paradox is obvious. Platforms have built in generative tools to simplify content creation and boost engagement. They got exactly what they were bound to get: a flood of low-quality generated content. Now they’re introducing filters, detectors, and report buttons—but they’re drawing the line so that their own AI tools remain on the “right” side. The question that no one is seriously addressing yet is: what should be done with content where AI was used 70% of the time? Or 30%? There is no answer, and it is precisely this gray area that will be the main battleground.

Source: Hacker News - Newest: ""AI" "LLM""