This week, Snapchat stopped rewarding fully AI-generated videos in its Spotlight creator programme. Substack launched a tool to help readers identify AI-written newsletters. LinkedIn added a “Seems like AI slop” report button. YouTube tightened monetisation rules around inauthentic and template-based content. Meta removed an Instagram AI photo-editing feature following user backlash.
Five platforms. One week. Same direction.
If you use AI in your content work, the reaction you might have is: should I pull back? The short answer is no. But the longer answer is worth understanding, because the crackdowns are far more specific than the headlines suggest.
The distinction that actually matters
Snapchat’s rule is worth reading carefully. It bans fully AI-generated videos from Spotlight monetisation. AI-enhanced content, human footage with AI editing, AI effects, AI voiceovers layered onto real recordings, is still eligible. The line is between content where a human made the creative decisions and content where the AI made all of them.
Substack’s flagging tool and LinkedIn’s report button follow the same logic. They are designed to surface content with no meaningful human editorial judgment behind it. A newsletter that is a ChatGPT output with a send button is what they are targeting. A newsletter where you used AI to research, structure your thinking, and polish your prose, and then edited and published it yourself, is not.
This distinction matters because two common responses to this news are both wrong. “Platforms are banning AI” is wrong. “AI content is safe as long as you add a human sentence at the end” is also wrong. What platforms are actually responding to is the absence of genuine human judgment in content, not the presence of AI tools in the process.
The number Substack’s CEO cited
Substack’s co-founder cited research suggesting up to 40% of writing on social media is now fake or AI-generated. The sourcing on that figure is not fully transparent, so treat it as an indicator rather than a precise measurement. But the direction is not in dispute. Platforms are seeing it in their own data, which is why five of them moved in the same week.
What this means practically: the signal-to-noise problem online is bad enough that platforms are building infrastructure to manage it. Marketers who publish human-quality content consistently are going to benefit from these tools, not be hurt by them. The ones who built content operations on volume-first AI production are facing a platform squeeze.
What changes for marketers who publish regularly
Nothing changes if you are already treating AI as a writing tool rather than a writing replacement. Drafting with AI, editing with judgment, publishing with your name and perspective on it, that is still fine across every major platform.
What is worth reviewing:
- Social content going out at volume without a human reviewing each piece. Platforms are getting better at detecting template-generated posts, and automation that looks automated will be deprioritised in recommendations.
- Newsletters or email sequences where an AI wrote and you scheduled without meaningful editing. If those go under a personal name without genuine human voice, they will increasingly get flagged and filtered.
- Short-form video content built entirely from AI generation with no original creative input. This has been declining in organic reach for months. These policies accelerate that decline.
If your AI content strategy is “produce more things faster,” platforms are building friction into that model. If it is “think better, write clearer, and review everything before it goes out,” you are fine.
The uncomfortable part
Most marketing teams do not have an AI content problem. They have a review problem. The AI tool produces something, it goes into a queue, someone hits publish. The actual editorial judgment, is this worth saying, is this the right angle, does this sound like us, gets skipped.
That is the behaviour these platforms are responding to. The tool is not the issue. What happens between the tool output and the publish button is.
The kind of AI-assisted content that survives these platform shifts is content with a real editorial layer, where AI handles the mechanical work and a human handles the judgment. That is what mark8ng.ai is designed to support: not more volume, but better decisions before you hit publish.