More than 70% of Musk's Twitter fake followers? Twitter's 44,000 active accounts were sampled by a data analysis company

 　　Tesla CEO Elon Musk's deal to buy Twitter is currently "on hold". According to Musk, he remains committed to the acquisition and just needs Twitter to clarify the issue of its fake accounts reaching 5%.

　　Recently, a number of new independent agencies have counted fake accounts on Twitter, and according to statistics in cooperation with SparkToro and other agencies, the proportion of fake accounts on Twitter is about 20%. It is worth noting that the proportion of fake accounts among fans of public figures is very high. For example, about 50% of fans of US President Biden are fake accounts, while the proportion of fake accounts of Musk fans is as high as about 70%.

　　On May 17, 2022, the US "Newsweek" reported that according to statistical tool data, 49.4% of the follower accounts of US President Biden's official Twitter account "POTUS" are fake accounts.

　　Of the 22.2 million followers of POTUS, about 10.9 million followers are fake accounts, accounting for 49.3% of the total followers. According to the statistics, among accounts with tens of millions of followers, the median proportion of fake account followers is 41%. That is to say, the proportion of "fake followers" among the Twitter followers of the US president is relatively high.

　　The statistic also noted that of the 34 million followers of Biden’s personal Twitter account, “Joe Biden,” fake accounts accounted for a relatively small proportion, with about 14.7 million fake accounts, or 43 percent.

　　Among the 93.39 million followers of Musk's Twitter account, the proportion of fake accounts may exceed 70%, or about 65.56 million.

　　And Musk himself has admitted that even if his tweets are hotly debated, the highest number of likes or comments is only 4.8 million.

　　After this news was announced, Musk also tweeted to comment: "So (just use a calculator), the number of fake followers of President Biden is 10 times that of the 5% that Twitter claims, or according to my In terms of the proportion of fake followers, it is 14 times that of 5%.”

　　However, their definition of fake accounts in this statistics is different from Twitter’s definition of fake accounts.

　　On the one hand, Twitter's definition of "daily active users" is any individual, organization or person who has ever logged in to Twitter through the Twitter APP or other paid products, advertising platforms, or any other means on any day and has been authenticated. For other types of accounts, this definition is also used in the company's statistics for the first quarter of 2022.

　　However, Twitter has not publicly disclosed its definition of fake accounts or a full method of how to classify them.

　　On the other hand, fake accounts are defined as those that do not regularly participate in writing and posting tweets, conducting activities on the timeline, and participating in various activities in the Twitter ecosystem.

　　When making statistics, randomly select 2,000 users from the 100,000 users who have recently followed an account for a sample survey. The tool then conducts an analysis of the sampled 2,000 accounts, including correlations between these accounts and more than 25 factors or indicators related to "spam/bots/low-quality accounts," etc.

　　Although none of the 25 factors or metrics pointed to fake accounts individually, an account was more likely to be fake when it was correlated with multiple metrics.

　　Of the sampled accounts, 81% had followers made up of new accounts created in the past 90 days; 93% of accounts had links to profiles that were fake links; 91% of accounts had only a "few" suspicious followers”; 88% of accounts only “follow very few other accounts”; 24% of accounts use the system default avatar; 45% of accounts have account names associated with spam; 72% of accounts have tweets The text contains spam-related content.

　　In its algorithm analysis of counting fake accounts, it also wrote that according to the agency's definition of fake accounts, some "fake accounts" may not necessarily be problematic. For example, there are news accounts that post top news stories through bot summaries, or travel accounts that use artificial intelligence to post photos and links to restaurants around the world.

　　Nonetheless, most of the accounts identified as fake by the agency share some common behaviors, including bad advertising, posting false information, promoting online viruses or malware, manipulating stocks and cryptocurrencies, and attempting to harass other users. .

　　Their analysis may have misclassified active users who don't tweet and only view other people's accounts as fake accounts, and may have missed some more sophisticated spam accounts. Its analysis of the proportion of fake spam accounts is a relatively "conservative" estimate.

　　In addition, through the statistics of more than 44,000 active accounts in the past 90 days, 19.42% of the accounts may be fake accounts. Through the statistics of more than 500,000 accounts in the past three and a half years, 22.35% of the accounts may be fake accounts.