Beyond the 90-9-1 Rule: How Algorithmic Feeds Concentrated Internet Attention

On May 6, 2026, a team of computer scientists published an analysis of 2.65 million TikTok videos, mapping exactly how human attention distributes itself when an algorithm decides what plays next. They found a brutal Pareto distribution: 20% of the videos accounted for 80% of the total views. The data, collected from a global donation program of 100 users, proved that push-based recommendation systems create a winner-take-all environment for visibility.[2]

This finding finalized the death of a foundational law of the internet. For two decades, digital culture was defined by a participation inequality metric coined by usability researcher Jakob Nielsen. "In most online communities, 90% of users are lurkers who never contribute, 9% of users contribute a little, and 1% of users account for almost all the action," Nielsen wrote in 2006. In the era of forums, blogs, and early Wikipedia, the barrier to creating content was high, leaving a tiny minority to generate the material that everyone else consumed.[1][3][6]

The social graph era of Facebook and Twitter slightly softened this participation inequality, but reach remained strictly gated by follower counts. If a user had no followers, they had no voice. The algorithmic feed—pioneered by TikTok and aggressively adopted by Meta across its platforms in late 2022—promised to democratize this system. By decoupling distribution from the social graph, the new architecture meant anyone's first video could theoretically reach millions based purely on engagement metrics.[4]

But while the barrier to creation plummeted, the concentration of attention skyrocketed. When an algorithm relentlessly optimizes for aggregate watch time, it heavily biases toward proven, highly-engaging content. The Gini coefficient—an economic metric where 0 represents perfect equality and 1 represents a total monopoly—reveals the scale of this shift. For context, the most unequal national economies on Earth sit around a Gini coefficient of 0.63.[4]

Algorithmic feeds blow past those economic limits. Data from StreamWrapped, which tracked 22,265 TikTok Live streamers between May and June 2026, found a Gini coefficient of 0.97 for virtual gifting. The top 1% of creators captured 70% of all revenue, while the bottom 50% earned an average of just 28 cents each. The attention economy has reached a level of inequality that physical economies cannot sustain.[5]

The mechanism driving this concentration is "lookahead caching." Unlike YouTube, where users actively select videos, algorithmic feeds push a continuous stream. As the May 2026 arXiv study detailed, platforms load a "manifest file" of upcoming videos into the user's device in advance. To keep users swiping and reduce server costs, the algorithm serves the exact same viral hits to millions of overlapping users simultaneously, maximizing temporal overlap and minimizing cache misses.[2]

This creates a paradox at the heart of modern internet culture. The feed feels infinitely diverse because the creators of the viral hits rotate rapidly, creating an illusion of a thriving middle class. A user might see ten different unknown creators go viral in a single day. But mathematically, the attention is hyper-concentrated on whatever handful of videos the algorithm is currently pushing to the top of the Pareto distribution.[2][4]

The 90-9-1 rule described a world where participation was scarce but attention was decentralized across thousands of niche forums. The algorithmic era inverted the equation: participation is abundant, but attention is a monopoly. The structural fix to this concentration will not come from tweaking recommendation weights, but from the ongoing rollout of decentralized protocols like AT Protocol, which allow users to choose their own algorithms and break the single-feed monopoly entirely.[4]