As Elon Musk has been publicly and loudly deliberating about the percentage of inauthentic accounts on Twitter ahead of a potential $44 billion acquisition of the company by the Tesla and SpaceX chief, Israeli researchers have placed the number in the double digits.
According to an independent analysis performed by Israeli tech startup Cyabra, the developer of a software platform that tracks fake users and accounts on social media for disinformation purposes, Twitter’s inauthentic users stand at just over 13%.
Since announcing his $44 billion offer to buy Twitter last, Musk has placed the issue of fake accounts in the limelight, saying he hopes to tackle it as part of an overhaul when/if he takes over. Last week, he put the takeover bid “temporarily on hold” as he suddenly sought more information on the problem. Musk said the hold was pending details of Twitter’s calculation that fake accounts were less than 5% of its users.
On Tuesday, he said the deal to buy Twitter could not move forward unless the company shows public proof that the figure was less than 5%. A day earlier, he gave the strongest hint yet that he would like to pay less for Twitter than his $44 billion offer and that, according to his estimates, at least 20% of Twitter’s 229 million accounts were spam bots, a percentage he said was at the low end of his assessment.
Musk told a Miami technology conference that a viable deal at a lower price would not be out of the question, according to a report by Bloomberg News, which said it viewed a livestream video of the conference posted by a Twitter user.
His appearance came hours after he began trolling Twitter CEO Paraj Agrawal, who posted a series of tweets explaining his company’s effort to fight bots and how it has consistently estimated that less than 5% of Twitter accounts are fake.
These events bolstered theories from analysts that Musk either wants out of the deal or is seeking a lower price, largely due to a huge decline in value of Tesla stock, some of which he has pledged to finance the Twitter acquisition. Twitter shares closed Monday down just over 8% at $37.39, below where the stock was just before Musk disclosed that he was Twitter’s largest shareholder. Musk made the offer to buy Twitter for $54.20 per share on April 14.
Speaking to The Times of Israel Tuesday, Cyabra VP of marketing Rafi Mendelsohn said that according to the company’s research, Twitter’s fake users amount to 13.7% of accounts.
Mendelsohn said Cyabra “looked at numbers for the last 12 months, so from May  to May , to get this number. We looked at millions of accounts.”
The Tel Aviv-based company, founded four years ago, works with governments like Israel and the US, and brands such as Warner Media, to help them track disinformation efforts and detect fake accounts on social media platforms. These include Facebook (Meta), Twitter, Instagram, TikTok, Russian platform VK, and Telegram’s public groups, Mendelsohn said.
The good, the bad, the fake
Cyabra’s platform, which Mendelsohn described as a “social media search engine,” is driven by machine learning algorithms that can yield insights into conversations to detect disinformation, and determine influence and reach.
Mendelsohn said the tech allows us to see “the good, the bad, and the fake” on social media platforms.
Cyabra’s research in February, ahead of the Russian war on Ukraine, saw a sudden and dramatic increase in anti-Ukrainian content in the days immediately before the invasion on February 24. On Valentine’s Day, for instance, the number of anti-Ukrainian posts created by the sample of Twitter accounts jumped by 11,000% compared with just days earlier. Analysts believe a significant portion of the accounts are inauthentic and controlled by groups linked to the Russian government.
“When you see an 11,000% increase, you know something is going on,” Cyabra CEO Dan Brahmy said at the time.
Mendelsohn explained that the reach and influence of fake accounts can be a serious issue. As part of the Ukraine-related research, Cyabra “uncovered a community of fake accounts based in Poland tweeting anti-Ukraine content in Polish, it was a few dozen” and that is enough to spread very widely.
There are a number of different signals that could suggest a fake account, Mendelsohn said. “It could be the name, [or the handle,] it could be the picture, the tagging, the timing; like if an account is quiet most of the day then spends one hour posting like 58 tweets.”
He said the company then applies “authenticity scores” to the social media profiles to determine their status.
The fake accounts problem is one “social media companies won’t or can’t really do anything about,” Mendelsohn said. “Every account is monetized, [platforms are] making money on accounts, so it may not be in their interest to shut them down, and it is getting harder.”
“The level of sophistication is incredible,” he said. “The [disinformation] campaigns are very targeted, very sophisticated.”
“We are all being impacted by this sophistication. The tools created on the ‘bad side’ are very advanced,” said Mendelsohn. The “good side” needs the tools to counter these efforts, he added.
Disinformation campaigns could have serious consequences, Mendelsohn said, “like on share prices [of public companies], on elections, on products, or reputations.”
“Social media is like the town square and if some information is fake, that’s a problem.”
When Musk put the Twitter fakes at 5%, Cyabra thought “we could actually uncover this,” said Mendelsohn.
In a tweet Tuesday, Musk said that “20% fake/spam accounts, while 4 times what Twitter claims, could be much higher. My offer was based on Twitter’s SEC filings being accurate.”
“Yesterday, Twitter’s CEO publicly refused to show proof of 5%. This deal cannot move forward until he does,” he added.
In tweets on Monday, Agrawal acknowledged Twitter isn’t perfect at catching bots. He wrote that every quarter, the company has made the estimate of less than 5% spam. “Our estimate is based on multiple human reviews of thousands of accounts that are sampled at random, consistently over time,” Agrawal wrote.
Estimates for the last four quarters were all well under 5%, he wrote. “The error margins on our estimates give us confidence in our public statements each quarter.”
Musk, who has said bots plague Twitter and that he would make getting rid of them a priority if he owned the platform, responded to that tweet by Agrawal with a poo emoji.
“So how do advertisers know what they’re getting for their money?” Musk tweeted in a subsequent response about the need to prove Twitter users are real people. “This is fundamental to the financial health of Twitter.”
Twitter has put the under 5% estimate in its quarterly filings with the Securities and Exchange Commission for at least the last two years, well before Musk made his offer last month.
But in the filings, Twitter expressed doubts that its count of bot accounts was correct, conceding that the estimate may be low.
Musk had described his motivation for buying Twitter stems from a desire to ensure freedom of speech on the platform and to boost monetization of a website that is massively influential but has struggled to attain profitable growth.
He has also said he favored lifting the ban on Donald Trump, who was kicked off the platform in January 2021 shortly after the former US president’s efforts to overturn his election defeat led to the January 6 assault on the US Capitol.
Agencies contributed to this report.