New AI Predicts Which Posts Will Go Viral Before They Do

Every day, social media posts explode or fizzle out, often for reasons that seem random. But researchers may have found a way to take some of the guesswork out of virality. They created a new AI system called TFM4POP that predicts how widely a post will spread across a social network. Rather than just counting shares, it looks at several different types of information at once — the text, the images, basic facts like the time of posting, and the pattern of who shares what over time.

The tricky part, the researchers say, is that earlier tools only focused on one piece of the puzzle, like how many people shared a post in the first hour. That misses a lot. TFM4POP instead borrows a trick from "tabular foundation models" — AI systems trained on huge amounts of spreadsheet-like data. That lets it combine different kinds of information the way a smart analyst might, but far faster and at internet scale. It also tracks the whole life of a post, not just the first few minutes.

The system was tested on several real-world datasets and consistently beat other state-of-the-art methods at guessing which posts would catch on. It also does this efficiently, updating only a small fraction of its internal settings compared to older methods, which means it's cheaper and faster to train. That's important for any company that wants to use it in real time.

So what does this mean for you? If this technology reaches platforms and marketing tools, you could see smarter recommendations, better content moderation, and more effective campaigns. But don't expect perfect predictions — the paper itself notes that social media is unpredictable. Still, knowing the odds is far better than guessing.