One of my most popular posts in the past week was a piece on StatusPeople’s faker tool. Many of the comments had to do with how accurately the tool represents the percentage of fake and inactive followers in a Twitter stream and why celebrities seem to have a disparate proportion of fake accounts.
It occurred to me that for all of the discussion generated around purchased followers that I didn’t understand the exact mechanism by which fake accounts were generated and how vendors on Fiverr and other sites could offer thousands of Twitter followers for a negligible amount of money. I assumed that there is little set-up cost and that there is little labor necessary to generate and maintain these accounts, but beyond that had no idea. As it turns out, this information is hidden in plain sight.
What constitutes a fake Twitter account?
Before I get ahead of myself, let’s examine what constitutes a “fake” Twitter account. StatusPeople and other agencies have specific criteria but it’s in no way infallible. If it were, Twitter could simply apply those filters to its accounts and all spam would be gone. That said (for reasons that are evident below), some of these criteria are typically used to determine that a profile may be fake:
- Low follower / followed ratio
- No profile picture
- No headline
- No URL
- Diversity of access point (are they using more than one means to access Twitter be it mobile or third-party apps)
- Tweets within a certain threshold number
How fake Twitter profiles are generated
Software to generate thousands of Twitter accounts is widely available. Just search on YouTube. These programs send data to Twitter through proxy IP addresses, generate names and user names without the user going through Twitter at all (some will even scrape the information and pictures from random accounts and replicate profiles with a couple of random misspellings). So, when I talk about the ingeniousness of many developers using the Twitter API, I have to include the developers who devised these programs…. even though they diminish the Twitter experience and are using their genius for dark purposes.
Here is a random YouTube clip of one of these programs in action:
After the account is created, there are other software solutions that will help to manager followers and messages all from this third party interface. Software like this fuels the market for both Twitter followers and retweets on YouTube, eBay, Fiverr and other websites.
Why do Celebrities on Twitter have so many fake followers?
One thing that many commenters on my post noted was the prevalence of spam on celebrity accounts. Adam Justice of Social Media Sun made the observation that following some celebrities is a tactic used in attempt to legitimize the bot account. Jason Ding of Barracuda Labs substantiated Adam’s point in research that he did by buying followers and then analyzing the accounts that followed him.
Ding went on to point out that there was a qualitative difference in how sophisticated the follower accounts are created which justifies different pricepoints. For example a thousand accounts with random generated names, no headlines and picture of an egg would cost much less than a thousand accounts with profiles, headlines and a picture. There is software for either.
What we can reasonably determine
So if a tool like StatusPeople’s Faker app finds followers within the range that they consider “fake” – is that indicative that someone purchased followers? You can’t definitively say that. Antony Francis of Head of Lettuce sent me a blog post explaining how their site was hit with a surplus of 10K spammish Twitter followers, and how they analysed these accounts with the great SocialBro. So, just because they are present doesn’t mean they were bought.
For celebrities, since they are high-value targets of Twitterbots – their presence probably signals that the celebrity’s audience isn’t nearly as large as their Twitter following would indicate.
I almost think we should make a tacit assumption that politicians on a national stage buy Twitter followers. Gawker ran a piece during the Republican primary analyzing the Twitter following of all of the candidates, and found that all of them had a population that fell within a criteria indicating that they are likely fake (for Newt Gingrich it was about 92%) . And for all of the people who have given me the business for saying that Barack Obama bought Twitter followers, you can read in the article how his following breaks out as well.
In Gawker’s article they also make a really shrewd point that Sarah Palin was (and I’m sure still is) the most popular person when compared to Republican candidates in the primary. Her Twitter following was about 600K when they wrote that piece, much less than the candidates. She also had the least number of fake accounts associated with hers. In high school, I had braces and played in the band. While the popular kids had an entourage at lunch, I tagged along with whoever stuck around campus. Social media is same way. If you’re a celebrity you should have more followers than the average person. If you’re an average person and you have a humongous Twitter following – you must be really popular.
But if you’re an average person with a huge following for no apparent reason, someone may discover that you are more popular with bots than people and question why.