Why Fake News Spreads Faster on X — And the Simple Fix

X (formerly Twitter) publishes the code behind its recommendation system, and two researchers used it to answer a question we've all wondered about: why does false information spread faster than the truth? Their answer is surprisingly simple. The algorithm gives every type of reaction — a like, a retweet, a reply, a quote — its own predicted score, then adds them all into one final number. Whichever post scores highest gets shown to the most people, over and over.

Here's the problem: that math treats a quick like as just as meaningful as a thoughtful reply. Fake and misleading posts are very good at grabbing fast, cheap reactions. So they rack up points and get recommended again and again, even when almost nobody is actually discussing them seriously. Accurate, nuanced posts tend to earn slower, deeper engagement — fewer instant clicks, more real conversation — so they lose the ranking battle.

The researchers rebuilt the algorithm on a real dataset of X posts from the 2024 US election and ran simulations. Tweaking the weights — making likes count less, replies more — barely helped, and sometimes made things worse. But one rule change worked: only amplify a post once it's predicted to earn thoughtful engagement, like replies or quotes. They call it a "reflective-threshold gate" — a bouncer that waits for real conversation before letting a post into the main room.

Across 46 different test scenarios, this gate pushed low-credibility content out of the spotlight without reducing exposure for mainstream, reliable posts and without killing engagement overall. In plain terms: you can slow the spread of falsehoods on social media without making the platform boring, and without the company losing clicks. That's a rare finding — a fix that costs nothing and works.