New AI Finds the Fewest People Needed to Reach Everyone Online

Imagine you want one message to reach every person in a country of millions. You don't need to tell everyone directly — you need the right few people to pass it on. Finding that smallest possible group is a famous math puzzle that becomes practically impossible to solve as a network grows. Now, researchers have built an AI that finds a near-perfect group anyway. It's a "graph neural network" — AI that understands how people are connected — trained on 12,000 made-up networks, and it never needs answer keys to learn.

The speed is the headline. Their method runs up to 55 times faster than traditional optimization tools and 14 times faster than earlier AI that learned from labeled examples. In practice, that's the difference between waiting hours and waiting seconds — and it still works on networks the AI has never seen before. That matters when you're planning a campaign across millions of accounts.

Why should you care? The same tool powers viral marketing, political campaigns, emergency alerts, and public health messaging. A city fighting an outbreak could find the best local voices to amplify accurate information, cutting cost and delay. Unfortunately, it cuts both ways: the exact same technique helps bad actors spread misinformation more efficiently, or helps someone map out who to watch or pressure.

The honest catch: this is a preprint, not yet peer-reviewed, and it was tested on standard network datasets rather than live platforms. Real social networks are messier, constantly shifting, and harder to measure. Most importantly, reaching someone isn't the same as convincing them — exposure to a message doesn't guarantee anyone acts on it.