Just like fake YouTube views, the fact that there are fake Twitter followers shouldn't be a surprise to anyone. Companies, celebrities, and politicians have plenty of reasons to buy fake followers and it seems that many do.
A couple of Italian security researchers, Andrea Stroppa and Carlo De Micheli, are following up on earlier research with actual names of Twitter accounts that bought tens of thousands of followers in one go.
The first tell-tale sign that an account has acquired a fake following is a huge surge in popularity in a short amount of time.
But that alone can be accounted for by other means; for example, the account may simply be going viral. An even bigger sign that something is up is a sharp drop in follower count, after a similarly sharp increase.
This could happen when accounts stop paying for the fake followers, in which case the operators of the fake accounts network will send a mass unfollow command, or more likely when Twitter removes huge numbers of fake accounts in one move.
The researchers listed brands like Pepsi, Mercedes-Benz, and Louis Vuitton, and people like Newt Gingrich, US Representative Jared Polis, Dmitry Medvedev, the Russian prime minister, 50 Cent, and Sean Combs, aka Diddy.
All of these accounts – and more – saw a wild fluctuation in follower count that can't be explained by circumstances.
Diddy, for example, who tweets under the @iamdiddy account, added 185,399 followers in one day in June last year, 30 times more than his average daily gain.
But, last month, he lost 393,665 followers in one day, 65 times more than in an average day. The same goes for 50 Cent, who lost 190,342 followers in one day.
Pepsi saw a spike of 71,686 followers in one day in November 2011 – coincidentally or not, just enough for Pepsi's follower number to surpass Coca-Cola's.
Pepsi argues that the jump is due to a promotional campaign on Twitter and that they are real followers, not fake accounts. However, a similar campaign last year didn't have anywhere near the same effect on follower count for Pepsi.