In late 2024, AI-generated articles briefly overtook human-written ones for the first time, crossing 50% of new English-language web content according to an analysis of 65,000 articles by SEO firm Graphite, and that threshold has held through 2025.¹ This is interesting not just because it marks an inflection point in content volume, but because it crossed that threshold at the same moment that social platforms are routing their views based not on who you follow, but on how you behave.² Two systems converging at once. One produces the content. The other decides who sees it.
The human, in both cases, is no longer the source.
Why This Matters
The social graph is functionally dead. Early platforms distributed what your network said. TikTok, which now shapes media behavior across the industry, operates almost entirely on behavioral signals like watch time, replays, and scrolling patterns rather than social connections. Follows are a secondary input.² Instagram has moved in the same direction. Meta credited its 2023 user rebound not to network effects but to what it called its "AI discovery engine."³ The network of people you chose stopped mattering as much as the pattern of how you react.
AI is already producing the content those signals optimize. More than half of new web articles are now primarily AI-written, up from roughly 5% before the release of ChatGPT.¹ TikTok has labeled over 1.3 billion AI-generated videos on its platform.⁴ There is a loop closing with AI where the system generates content, humans react, the system learns which reactions to maximize, and the system generates more targeted content. The human is a sensor inside that loop, not the author of it.
What optimizes for engagement produces structurally different outputs than what optimizes for expression. Author-driven systems produce ideas, disagreement, curiosity, and the unexpected. Things people want to say. Things they feel. Optimization systems produce emotional triggers, behavioral consistency, and content that feels familiar because it was built from patterns of what previously worked. Those aren't the same content from an alternative angle. They are different things being produced for different effect. One is made for yourself. The other is made for an audience.
We are not watching a shift in how content gets created and distributed. We are watching a shift in what content is, and where it comes from.
It Happens Slowly, Then Suddenly
Think about it in terms of music streaming, where platforms like Spotify already do something similar. Artists still write music, but algorithms heavily shape what anyone ever hears. Now imagine the system also generated the songs. You listen. The system tracks your reactions. Then it generates a slightly different track tuned to your taste. You're no longer discovering music. The system is manufacturing it around your preferences. Social media platforms are already partway down that path.
🐜From biology: In ant colonies, most ants aren't making decisions about the colony's direction — they follow local signals, primarily pheromones, and large-scale behavior emerges from those tiny reactions. The structural parallel to large algorithmic platforms is imperfect but suggestive: individuals react locally, and the system evolves from the accumulated signals — without any single participant understanding what it's building.
Reaction
Social media was always a strange name. The "social" implied connection between people; the "media" implied expression. Both assumptions are now under pressure from the same direction, and the content inflection point has already happened. What comes next is a cultural question: whether people accept the shift from author to sensor, or whether a parallel layer of genuinely human-authored spaces emerges in response. That tension may decide where the internet goes from here.
Sources
¹ Graphite, "More Articles Are Now Created by AI Than Humans," https://www.graphite.io/five-percent/more-articles-are-now-created-by-ai-than-humans
² TikTok, "How TikTok recommends videos," https://newsroom.tiktok.com/en-us/how-tiktok-recommends-videos
³ Mark Zuckerberg, Meta Q4 2022 earnings call, February 1, 2023. Via Deadline: https://deadline.com/2023/02/facebook-instagram-meta-mark-zuckerberg-1235247083
⁴ TikTok Newsroom, "More ways to spot, shape and understand AI-generated content," November 2025. https://newsroom.tiktok.com/en-us/more-ways-to-spot-shape-and-understand-ai-generated-content