Some people do it just out of simple competition, essentially throwing their money away so they can boast more Twitter followers than their friends. Others do it to boost their corporate profiles, while even more high-profile cases have led to better reputations in the world of online clout, and thus job opportunities and advertising revenue.
Regardless of motivation, the fact remains that money is changing hands in order for certain Twitter accounts to increase the number in their "followers" column. And, a deep network of developers and merchants has arisen to feed the market, constantly looking to garner more money while dodging Twitter's vigilant eye.
Jason Ding is a research scientist at Barracuda Labs who worked on the company's recent investigative study into the economy behind fake Twitter accounts. He says people are selling fake followers out in the open, on eBay and Google Shopping. The beneficiaries range from follower-crazed teenagers to private startup companies to even presidential candidates, Ding says.
In fact, the study found that Republican candidate Mitt Romney's Twitter account saw a 21 percent increase in followers in July. A full 23% of them have never sent a tweet and 10% have already been suspended by Twitter. Ding was quick to point out that this does not necessarily suggest that Romney's campaign was responsible for the increase. Those purchasing followers only need to provide a link to the Twitter account that will receive the new followers. That means anyone can purchase fake followers for anyone else, even with malicious intent.
"It can promote or improve someone, but also it can destroy someone when they report them to the public and they're exposed," Ding says.
Several other high-profile cases have come to light in the past few months, such as disgraced former ESPN.com columnist Sarah Phillips and USA Today sports gambling expert Danny Sheridan. Evidence uncovered by Deadspin suggested the former was involved for promotional purposes, while the latter claimed it was done by a discontented third party.
On the business side, it's easy to see why people are selling Twitter followers. The study found that the average dealer has the capacity to control as many as 150,000 followers at a time, sometimes more. Those who can control 20,000 fake accounts and can attract sales of $20 or more -- the going rate is 1,000 followers for a minimum of $18 -- stand to earn roughly $800 per day, according to Barracuda Labs. Keep in mind that very little of this work is manual; Ding says the dealers could easily control a system of botnets and set up a few software tools to automate much of the process.
Twitter, meanwhile, has begun cracking down on the population of robot accounts on the site. Earlier this month, the company announced a handful of new tools designed to reduce the amount of spam and abusive accounts on the site.
How, then, are people building and maintaining an inventory of robot Twitter accounts to meet the demand?
Twitter's API used to break Twitter's terms of service
Building an inventory of fake Twitter accounts isn't difficult for anyone with access to an Internet connection and a Web browser.
Fake Twitter accounts need some information - an image to use as an avatar, perhaps some bio content, maybe even some tweets. For merchants that need to create an inventory of 20,000 accounts quickly, this simply cannot be done manually.
Twitter provides its API at dev.twitter.com for anyone looking to create applications. Once an app has been created, the developer is given an access token, which Ding described as "basically a password for your application." That token can be obtained by simply logging in with a Twitter account, entering a name, a URL and a description of the app the user intends to create. The access token enables the recipient to create an app that connects to Twitter's servers, where all information made public by the site's users is stored, Ding says.
Using Twitter's API, developers can design programs that collect all the information of a given group of Twitter users, such as, for example, the 800,000 users following Mitt Romney's account. These programs don't necessarily hijack these accounts -- they copy the images and text from their profiles and tweets. This pool of information can then be automatically ported into accounts based on an algorithm that automates the registration process on a massive scale.
"Anybody can use this strategy to grab information from Twitter -- either Mitt Romney's followers, or his tweets, or what he's tweeted, or activity," Ding says. "If you're just browsing the Internet and you open a browser and see all of a user's followers, that's taking time. But if you used the API, you can in one moment borrow 300 people's information."
By now, though, even this process has been open sourced. First-page results on a Google search for "program to make fake twitter accounts" return this forum on Freelancer.com, titled "Software create fake twitter users jobs." There, several development projects are posted for bidding that request a program that grabs information from Twitter users' accounts.
"Looking to design a client based application for Mac OSX that scrapes all posted tweets from each individual user who are following a specific user," one example reads. "All scraped content compiles into a txt [sic] file that can be saved to the client end computer."
The intent for this project is not listed anywhere in the description. This information could conceivably be used for legitimate purposes, such as collecting and researching a particular user's tweets. It could also be used to populate the empty fields of thousands of fake accounts, making them look more legitimate.
Indeed, as Twitter turns on the heat on abusers, the appearance of legitimacy is becoming an essential consideration for those who persist.
Twitter's reputation may be on the line
On the same day that Barracuda Labs released its report, Ding says Twitter suspended around 39,000 fake accounts. However, given that Twitter had suspended about 12,000 accounts a day earlier, Ding says he doesn't believe the report had anything to do with Twitter's crackdown. Twitter declined to comment.
What may have prompted the sting on Twitter abusers and spam accounts was a new Twitter initiative concerning the presidential election. On Aug. 1, just days before announcing its new anti-spam tools and carrying out its sting on fake accounts, Twitter released the Political Index tool. On this entirely new site, Twitter will stream data in real-time based on all tweets that mention either Romney or Barack Obama. The index shows how many tweets mention the candidate in a positive or negative light, and gauges public opinion accordingly.
Given how easy it is to control networks of robot accounts, Ding says these figures could be altered by anyone with a vested interest in controlling the data.
"Keep in mind that many of the dealers can make a lot of fake accounts on Twitter to make some impact on this political campaign, so they can control these fake accounts and post some positive messages and some negative messages," Ding says. "So they do certain things to influence the overall index in certain ways."
The Political Index is just an example of the larger problems Twitter faces if it can't keep up with the growth of fake accounts, Ding says.
"If anybody can buy followers easily, if many people are doing that, then the overall trust on [the] social platform is decreased gradually," he says. "Next time we see people who have followers, you don't think that's real."
Right now, Twitter appears to be stuck in a game of cat and mouse with its abusers. Ding credits the company's recent efforts, declaring that "Twitter is moving fast on this case now." At the same time, he acknowledges that those looking to cash in on the market for followers will become more difficult to catch, putting a thorn in Twitter's side as it aims to become a national resource for reliable information.
"We are also seeing there are some really high-quality fake accounts which have many tweets and many followers and many following," Ding says. "In that case, Twitter will have to be much more indexed and [use] many more metrics to identify them."