Musk’s Questions About Twitter Bot Problem Spur Race for Answer

Twitter indicates in regulatory filings that the number of automated accounts is less than 5% of the total, but Musk has said he suspects that the number is much larger

By

Bloomberg — Elon Musk has said that his $44 billion purchase of Twitter Inc. (TWTR) hinges on the accuracy of one figure: the number of bots on the platform. That’s triggered a race for answers.

Twitter indicates in regulatory filings that the number of automated accounts is less than 5% of the total, but Musk has said he suspects that the number is much larger. If he can demonstrate that Twitter’s figures are inaccurate, the billionaire may attempt to use the information to get out of the deal or negotiate a lower price.

Though many outside estimates do put the figure above the 5% threshold, their assessments and methodologies vary. Andrea Stroppa, a former data consultant for the World Economic Forum and a veteran of scrutinizing online counterfeit goods, estimates that bot accounts have accounted for about 10% of Twitter’s global audience over the past nine years.

The rate rises to as much as 20% for some specific topics such as cryptocurrencies, the researcher said, and above 30% for accounts engaged in certain conspiracy theories.

The research firm Bot Sentinel, meanwhile, estimates that 10% to 15% of accounts on Twitter are inauthentic. That includes fakes, spammers, scammers, nefarious bots, duplicates and single-purpose hate accounts. Inauthentic accounts are more likely to tweet about politics, cryptocurrency, climate change and Covid-19 than less controversial topics like kittens and origami, Bot Sentinel has found. Cyabra, a research firm with a different methodology, puts the percentage of inauthentic Twitter profiles at 13.7%.

Twitter didn’t immediately respond to a request for comment.

Musk said this week that spam bots amount to at least 20% of Twitter’s total, and he asked whether the true figure could be as high as 80% or 90%. He tweeted out a meme on Thursday jokingly suggesting that Twitter is “all bots.”

But it may be hard to use the issue to get out of buying Twitter, said Adam Badawi, a professor of law at the University of California at Berkeley.

“To have leverage, he would need to show that the difference between the true number of bots and the amount Twitter disclosed in its SEC filings rises to the level of a ‘material adverse effect,’” he said. “That’s M&A jargon for a complete calamity. He’d probably need to show that over half of Twitter’s users are not real.”

Musk, the head of Tesla Inc. (TSLA) and the world’s richest person, agreed last month to acquire Twitter and take it private for $44 billion. After that, as Musk sought to shore up financing for the transaction, he showed signs of wavering. He said last week that the deal was “on hold” until he gets more information, specifically proof from Twitter that so-called spam bots make up less than 5% of its users.

Earlier this week, Musk stoked speculation that he could seek to renegotiate the takeover, saying at a tech conference in Miami that a viable deal at a lower price wouldn’t be “out of the question.” But the transaction isn’t officially on hold or being renegotiated; Musk is still contractually committed to it.

Musk likely has a different experience with bots on the platform than most. Those designing automated accounts program them to follow popular users on a site, so that they fit in with the crowd and look more human. Musk, with a following of 94 million, probably attracts a higher percentage of bots than most users. His image has also been used by cryptocurrency accounts to run scams.

That said, bots are indeed more of an issue for Twitter than other platforms, Stroppa said.

“Twitter’s bot detection software is much weaker than other social media,” he said. On Twitter, you can run a botnet using outdated techniques and don’t need sophisticated tools such as Android emulators and 4G proxies, which can mimic customers using a mobile phone, Stroppa said. Professional botnets on Twitter are stealthy, especially the ones used for digital propaganda, he added.

Bots also aren’t against Twitter’s rules. Some are comedic or useful in other ways, such as one that automatically tweets out the magnitude of earthquakes.

With hot-button issues such as the pandemic, automated activity tends to be higher. Carnegie Mellon University researchers collected more than 200 million tweets discussing coronavirus or Covid-19 from January 2020 to June 2020. Of the top 50 influential accounts retweeting material, 82% were bots, they found. Of the top 1,000 retweeters, 62% were bots.

But the overall number may be close to unknowable, even by Twitter -- and it’s not even clear what constitutes a bot.

“I suspect that this will be drawn out in courts,” Duke University professor Chris Bail said. “Because reasonable people can disagree on what is or what isn’t a bot.”