One-Third of Trump Supporters on Twitter are Bots

Study Shows a Third of Trump Twitter Accounts are Bots

by Aya Katz

According to a study run by the Project on Computational Propaganda, nearly a third of active Twitter accounts that favor Donald Trump are bots. The study was run around the time of  the first presidential debate, and it also revealed that while 32.7% of pro-Trump tweets were from automated accounts,  22.3% of pro-Clinton tweets were from bots as well. The study did not address the presidential campaigns of Gary Johnson and Jill Stein on Twitter. It may well be that all those tweeting about them are actual people.

The abstract to the scholarly paper runs as follows:

Bots are social media accounts that automate interaction with other users, and political bots have been particularly active on public policy issues, political crises, and elections. We collected data on bot activity using the major hashtags related to the U.S. Presidential debate. In this brief analysis we find that (1) Twitter traffic on pro-Trump hashtags was roughly double that of the pro-Clinton hashtags, (2) about one third of the pro-Trump twitter traffic was driven by bots and highly automated accounts, compared to one fifth of the pro-Clinton twitter traffic, (3) the significant rise of Twitter traffic around debate time is mostly from real users who generate original tweets using the more neutral hashtags. In short, Twitter is much more actively pro-Trump than pro-Clinton and more of the pro-Trump twitter traffic is driven by bots, but a significant number of (human) users still use Twitter for relatively neutral political expression in critical moments.

Criteria used to determine whether a Twitter account was operated by a bot included the following:

• no profile picture
• no friends
• following more than followed
• a monomania for one subject
• never sleep
• post very frequently (500 times a day, for instance)
• reply to a message in less time than it takes to read it
• send same message to many others

The researchers labeled all tweets at the time of the first presidential debate as either pro-Trump, pro-Clinton, neutral to one or the other or both.

While there were many more pro-Trump tweets than pro-Clinton tweets, roughly a third of the pro-Trump tweets were from bots.

The researchers noted that there was considerable difference between the behavior of bots and regular Twitter users:

There is a noticeable difference between the usage patterns of typical human users and accounts that are bots or highly automated. For example, the top 20 accounts, which were mostly bots and highly automated accounts, averaged over 1,000 tweets a day and they generated 86,000 tweets. The top 100 accounts, which still used high levels of automation, generated around 200,000 tweets at an average rate of 500 tweets per day. In contrast, the average account in the whole sample generated one tweet per day.

Of course, it’s true that some human Twitter users are more passionate and vocal than others. Some tweet more frequently. But from the results of this study we can see that Trump and Clinton have far fewer supporters on Twitter than we have been led to believe. It’s quite possible that Johnson supporters are also much more numerous than Twitter traffic might indicate.