X Optimized for This. Now Look at It.

There's a version of this story where the algorithm is the villain. Mute, mechanical, indifferent — just doing what it was told. That version is too easy.

The harder version: someone decided what to tell it.

What the data keeps saying

Earlier this year, a researcher at Nieman Journalism Lab ran a fairly rigorous exercise — scraping the 200 most recent posts from 18 large publisher accounts on X and tracking engagement across each. The finding wasn't subtle. Accounts that aggregate breaking news without posting links dramatically outperformed traditional publishers who linked out to their actual reporting. The platform wasn't neutral about this. It had a preference, and the preference was: stay here, don't leave.

Around the same time, Nate Silver went looking at which accounts were actually winning on X — pulling the accounts with the most engagement through early 2026 and mapping them. His conclusion, as Daring Fireball noted, was that the top performers skewed hard right, yes, but there was a second observation Silver found equally important: the quality was low. Accounts he'd never heard of. Content he described as a freak show. Whatever X's algorithm was selecting for, it wasn't rigor.

Put those two findings side by side and you get something more specific than "Twitter went bad." You get a system that penalizes links and rewards pure engagement bait — and then you look at what floats to the top and realize those two facts are not a coincidence.

Engagement and utility were always going to fight

This is the part the platform never really resolved. Engagement — likes, comments, reposts — is not the same thing as usefulness. It never was. A headline that makes you furious generates more engagement than a story that makes you informed. An account that parrots a WSJ scoop without crediting it, without linking to it, without adding anything — that account wins on the metrics X apparently cares about.

The Nieman research captured this precisely: accounts like Leading Report, which post things like "BREAKING: Iran has halted direct talks with the US, per WSJ" — no link, just the claim — sit in a category the researcher specifically tracked. They're not publishers. They're aggregators. And they clean up.

The publishers who do the actual reporting? They post the link, they lose the reach, they fund the journalism that feeds the aggregators who eat their lunch on the platform that deprioritized them.

You can call this ironic. I'd call it a business model — just not one that benefits journalism.

What makes this worth sitting with isn't the rightward tilt or even the quality collapse, though both are real. It's that the platform had a choice — engagement or utility — and made it clearly, quietly, and in a way that took two separate independent analyses to fully articulate. Musk didn't announce it. The algorithm just got better at what it was actually being asked to do.

And now we have the results: a bubble chart full of accounts nobody's heard of, dominating a platform that once, in some earlier life, felt like the closest thing the internet had to a shared conversation.

The platform didn't accidentally become this. It was tuned.