Trump’s Got The Twitter Bot Vote Sewn Up

Unfortunately for Trump, Twitter bots can't vote

Oct 18, 2016 at 3:29 PM ET

Republican presidential nominee Donald Trump may lag behind his Democratic opponent Hillary Clinton almost every reputable poll, but there’s at least one area where he has a commanding lead: Twitter bots.

Automated accounts tweeted more than four times as many pro-Trump messages as they tweeted pro-Clinton messages during the first presidential debate, according to a study from an Oxford University professor Philip Howard’s Project on Computational Propaganda, which looks at bots on social media to try to determine if and how they may manipulate public opinion on politics, for instance by making a certain issue or candidate seem to have more support than it or he actually does. Bots, the study says, have “become a means of managing citizens,” and have had a “modest but strategic role in the U.S. presidential debates.” And Trump supporters appear to be much better at creating and amplifying their content on Twitter than the pro-Clinton camp.

Researchers collected a “sample” of 9 million tweets between September 26—the day of the first presidential debate—and September 29 that included certain hashtags identified as pro-Trump, pro-Clinton, or neutral—think #MAGA, #CrookedHillary, #ImWithHer, and #Debates2016—to identify the political leanings of those tweets. That means tweets that didn’t have those hashtags were not counted, and the tweets were collected using Twitter’s Streaming API, which gives a sample of tweets and not all of them.

Of those 9 million tweets, about half of them contained hashtags about the debates or candidates. Of those, researchers found that about 40 percent were pro-Trump, about 35 percent were neutral, and just under 15 percent were pro-Clinton. But when researchers looked at the accounts that were tweeting these messages, they found that Trump’s Twitter supporters were more likely to be bots than Clinton’s, and bots accounted for more than four times as many pro-Trump tweets as pro-Clinton tweets.

Researchers classified accounts that tweeted at least 50 times per day over the study’s four day period as having “high levels of automation.” They also identified bot accounts by simply searching for ones that claimed to be bots in their profile or in the tweets themselves. In all, bots accounted for 20 percent of the debate tweets, and the top 20 accounts were “mostly bots,” which accounted for 86,000 tweets over the course of the study. By contrast, the “average account” in the sample tweeted once per day.

While the volume of pro-Trump bot tweets was much greater than pro-Clinton bot tweets, the percentage of pro-Trump tweets sent by bots was closer to that of pro-Clinton tweets sent by bots. About a third of the nearly 1.8 million pro-Trump tweets came from bots (that’s about 575,000 in all). Twenty-two percent of the nearly 615,000 pro-Clinton tweets came from bots (or 135,000). And only 11 percent of 1.6 million neutral debate tweets were believed to come from bots (185,000). There were more human-authored pro-Trump messages than pro-Clinton ones as well — about 2.5 times as many.

Trump has more support on Twitter, then, from both humans and bots — or at least, he did during the first presidential debate (it’s worth noting that Trump was much closer to Clinton in the polls at that point, too. He isn’t anymore). Considering how Twitter has become somewhat of a haven for alt-right trolls — to the point that it reportedly lost two potential buyers because of its inability to deal with its users’ trolling and abuse — who tend to see Trump as their candidate of choice, this really isn’t a surprise. The question for Trump, then, is how many of them can and will actually vote for him.