Twitter says the average user with the algorithmic timeline follows 10%-15% more accounts than those who have reverted to the old reverse-chronological timeline
The algorithm hasn't fixed harassment, but it did remedy something else — t's hard to remember now, but there was a time … Tweets: @medium , @_harshilshah , @campuscodi , and @willoremus Tweets: @medium : Who—and how many people—to follow has been an issue for Twitter users since the beginning. One change, once considered “the end of Twitter,” fixed it. http://read.medium.com/MsvWLiZ Harshil Shah / @_harshilshah : I was following like 500 people back when I used Tweetbot and even that felt like a lot Got to 1100+ with the algorithmic timeline but ended up scaling it down a bit to just about a thousand now https://twitter.com/... Catalin Cimpanu / @campuscodi : @Techmeme @WillOremus The old algorithmic timeline surfaced more interesting content. The reverse-chronological timeline is great for journalists but probably pretty boring for regular users. Site looks dead when you're getting recent tweets, as there's not a constant stream of tweets, as you'd expect Will Oremus / @willoremus : There was a time when Twitter's most visible problem wasn't hate or harassment but stagnation. It still hasn't solved the former—far from it. But I wrote about some new data that helps to shed light on how it has partially solved the latter. https://onezero.medium.com/...
Context & Ripple Effects
Twitter's algorithmic timeline has been contested territory since the company was forced to let users revert to a chronological timeline in 2018 after the ranked feed was branded "the end of Twitter." This OneZero data point reframes that fight: the algorithm isn't just reordering tweets, it is quietly growing users' follow graphs — average algorithm-timeline users follow 10%-15% more accounts than those who switched back.
That matters because the follow graph is the raw input to everything else on the platform. A later study found the same algorithmic timeline greatly reduces exposure to external links versus chronological ordering, meaning ranking choices reshape both how much people follow and what they see from it.
First-order effects
- Users on the ranked feed are carrying 10%-15% larger follow graphs than their chronological peers, which means more accounts competing for placement in every ranked session — and more content the algorithm must curate per user.
Second-order effects
- A larger follow graph gives Twitter's ranking system more leverage over what each user actually sees, compounding effects like the measured suppression of external-link exposure on the algorithmic timeline.
Third-order effects
- Feed defaults function as social-graph policy: by the time Twitter's web client stopped defaulting to the algorithmic feed in 2023, ranking had already demonstrably reshaped who users follow — evidence that timeline design is one of the strongest structural forces on a social network's topology.
The trend: Timeline-ranking algorithms are becoming de facto architects of social graphs, shaping follow behavior as much as the content they reorder.