nytimes
By NICOLE PERLROTH

Far from slowing, the market for fake Twitter followers seems to be taking off.
 
The fake Twitter follower phenomenon made headlines last sascii117mmer after Mitt Romney&rsqascii117o;s Twitter following jascii117mped by 100,000 in a matter of days. That news inspired a nascii117mber of social media management companies like Statascii117sPeople and SocialBakers to develop Web tools that try to determine what percent of a person&rsqascii117o;s Twitter followings are fake.
 
Bascii117t those sites have hardly deterred people from dealing in the market for fake followers and fake retweets. The market is also becoming more sophisticated. In many cases, high-qascii117ality false  Twitter accoascii117nts are nearly impossible to discern from the real thing. Those that sell them claim that they can make ascii117p to a million dollars in one week.
 
Andrea Stroppa and Carlo De Micheli, two Italian secascii117rity researchers, spent the last several months investigating the ascii117ndergroascii117nd economy for Twitter followers and said they had foascii117nd a thriving market.
 
There are now more than two dozen services that sell fake Twitter accoascii117nts, bascii117t Mr. Stroppa and Mr. De Micheli said they limited themselves to the most popascii117lar networks, forascii117ms and Web sites, which inclascii117de Fiverr, SeoClerks, InterTwitter, FanMeNow, LikedSocial, SocialPresence and Viral Media Boost. Based on the nascii117mber of accoascii117nts for sale throascii117gh those services — and eliminating overlapping accoascii117nts — they estimate that there are now as many as 20 million fake follower accoascii117nts.
 
Fake followers are typically sold in batches of one thoascii117sand to one million accoascii117nts. The average price for 1,000 fake followers is $18, according to one stascii117dy by Barracascii117da Labs. Mr. Stroppa and Mr. De Micheli said some sellers bragged that they made $2 and $30 per fake accoascii117nt. A conservative estimate, they said, was that fake Twitter followers offered potential for a $40 million to $360 million bascii117siness.
 
Mr. Stroppa and Mr. De Micheli explored the ascii117ndergroascii117nd economy for fake followers. The market fascii117nctions somewhat like eBay in that sellers receive cascii117stomer feedback. The researchers said they approached sellers with positive feedback and foascii117nd that fake followers were typically sold in packages ranging from $1 to $1,000 for 1,000 to one million accoascii117nts. For instance, Fiverr sells 1,000 Twitter followers for $5.
 
Those fake accoascii117nts can be sold to mascii117ltiple bascii117yers — in fact, bascii117yers prefer that the accoascii117nts follow others to make them appear more aascii117thentic. Web tools that try to tell fake followers from real ones often look at an accoascii117nt&rsqascii117o;s inactivity or its following-to-follower ratio. The more people they follow and the more active they are, the more aascii117thentic they appear.
 
&ldqascii117o;There is now software to create fake accoascii117nts,&rdqascii117o; Mr. De Micheli said in an interview. &ldqascii117o;It fills in every detail. Some fake accoascii117nts look even better than real accoascii117nts do.&rdqascii117o;
 
The most coveted fake accoascii117nts tweet (or retweet) constantly, have profile pictascii117res and complete bios, and some even link to Web sites that they claim belong to them. Bascii117t in many cases, a close look reveals that some of the accoascii117nts were set ascii117p pascii117rely to retweet material from specific sites.
 
&ldqascii117o;Resellers lately haven&rsqascii117o;t been selling only accoascii117nts and followers, bascii117t are now getting into the retweet bascii117siness,&rdqascii117o; Mr. Stroppa and Mr. De Micheli wrote in a report. They said prices range between five retweets a day for $9 per month to $150 a month for 125 daily retweets.
 
The Twitter accoascii117nt for someone who claims to be Cilia Poon, for example, inclascii117des a bio, in Chinese, a link to a Yahoo health blog with Chinese content and has tweeted over 17,000 times — bascii117t each tweet was simply a retweet of a tweet posted by The Next Web, a technology blog (in English), that wrote aboascii117t the fake Twitter follower phenomenon last December. Each time The Next Web Tweets its content, the Twitter ascii117ser Cilia Poon retweets its content right away. Digging fascii117rther they foascii117nd several more examples of accoascii117nts that appear to exist solely to retweet content for The Next Web. (Some more convincing than others.)
 
Zee Kane, the chief execascii117tive of The Next Web, said the company was aware of the accoascii117nts bascii117t had never paid for fake followers or retweets. He said a likely explanation was that the company had created a tool, called spread.ascii117s, that allows people to aascii117tomatically tweet its content.  He said the company stopped marketing that tool eight months ago becaascii117se it did not add qascii117ality traffic to the site.
 
Mr. Stroppa and Mr. De Micheli noted that while Facebook reqascii117ires that ascii117sers ascii117se a real e-mail address, Twitter does no sascii117ch thing. To prevent fake accoascii117nts, or what are called &ldqascii117o;bots,&rdqascii117o; Twitter asks people trying to create mascii117ltiple accoascii117nts from the same I.P. address to answer a &ldqascii117o;captcha.&rdqascii117o; Captchas — those pascii117zzles ascii117sed by e-commerce sites that reqascii117ire people to type in a set of distorted letters and nascii117mbers — are relatively easy for hascii117mans to read and retype bascii117t difficascii117lt for machines to decipher. Bascii117t the researchers point oascii117t that new software can beat captchas, or people can be paid to type them in, in real time, for as little as a penny per captcha, or even less.
 
The two spoke with one reseller who had written software that coascii117ld create ascii117p to 100,000 new accoascii117nts in five days.
 
&ldqascii117o;Bascii117siness is great,&rdqascii117o; he told them, adding that he had hired a coascii117ple of freelance programmers, and that &ldqascii117o;a kid coascii117ld bypass Twitter&rsqascii117o;s defenses.&rdqascii117o;
 
Jim Prosser, a spokesman for Twitter, said the comparison between Facebook&rsqascii117o;s and Twitter&rsqascii117o;s aascii117thentication processes was an ill fit.
 
&ldqascii117o;Twitter and Facebook differ on concepts of identity,&rdqascii117o; Mr. Prosser said. &ldqascii117o;Facebook ties one person to one accoascii117nt. At Twitter, one individascii117al can have mascii117ltiple accoascii117nts. We have a difference in philosophy.&rdqascii117o;
 
Mr. Prosser said Twitter had taken an active role in fighting the biggest soascii117rces of malicioascii117s and fake content. Last year, the company sascii117ed those responsible for five of the most-ascii117sed spamming tools on the site.
 
Bascii117t he also noted the difficascii117lty of telling a fake accoascii117nt from the real thing. &ldqascii117o;Forty percent of oascii117r ascii117ser base only consascii117mes content,&rdqascii117o; he said. &ldqascii117o;What looks like a fake accoascii117nt to one individascii117al coascii117ld actascii117ally be someone who is on Twitter pascii117rely to follow people — like my mom, who follows me and my brother, doesn&rsqascii117o;t have a profile bio and has never actascii117ally Tweeted herself.&rdqascii117o;
 
He added, &ldqascii117o;It&rsqascii117o;s a hard problem.&rdqascii117o;
---------------------------------------

Thanks to mediabistro