Scientists Built a Way to Measure How Much Bots Really Sway Twitter

Social media platforms have gotten decent at spotting coordinated fake accounts — the bot networks and troll farms that push talking points in unison. But spotting them has never answered the obvious next question: how much do they actually change what people see and believe? A team of researchers from Italy and Japan just published a way to answer that, using math rather than guesswork.

Their approach works in two ways. First, they built a method that treats a viral conversation like a tree of shares and calculates the best possible placement for fake accounts — essentially asking, if these bots were perfectly positioned, how far could their message travel? Second, because real platforms rarely share that kind of detailed data, they created a simpler method that estimates influence using retweet patterns alone. Think of it as a weather forecast versus a full climate model: one is easier to run, but they checked that both give similar answers.

They tested this on tweets from the 2019 UK general election and on documented state-backed influence campaigns from several countries. The results were mixed and interesting. In the UK election data, coordinated accounts had surprisingly limited reach. But across the state-backed campaigns, the picture flipped: several operations spread their messages about as widely as the most well-connected, genuinely popular accounts on the network. That's the finding worth paying attention to — some propaganda machines aren't inefficient at all.

The researchers also checked whether their own method was skewing the results, and it wasn't. The effects they measured came from how these networks are structured, not from quirks in the math. In plain terms: this gives election watchers, journalists, and platforms a ruler for measuring how loud a disinformation campaign actually is — so they can spend limited attention on the operations that matter most.